Macrofungi signify an extensive reservoir of bioactive metabolites. Still, their chemical description is stalled by fragmented analytical pipelines and low reproducibility. This study introduces SysFungiNet, a unified, FAIR-compliant systems-bioinformatics framework designed to accelerate bioactives discovery in non-model fungi. Unlike existing pipelines, SysFungiNet integrates LC-MS/MS metabolomics, transcriptomics, and genomic data with explainable artificial intelligence and molecular docking validation. By reconstructing biosynthetic pathways for Ganoderma lucidum and Craterellus cornucopioides, the framework achieved a Pathway Completeness Index of 0.86. An ensemble machine-learning model predicted bioactivity with an F1-score of 0.91, identifying 312 annotated metabolites. Crucially, decision-making was transparent, with SHAP analysis identifying specific chemical substructures driving predicted immunomodulation. The framework prioritized five high-confidence candidates, including a putative novel terpenoid, Cornucopiolide, which showed high binding affinity (−9.4 kcal/mol) to human immune receptors in silico. Benchmarking against MetaFungi and PhytoOmics demonstrated superior annotation accuracy and reproducibility. SysFungiNet offers a scalable, open-source ecosystem for transforming fungal bioprospecting into an evidence-driven process.

Fatigue-related impairment is a major contributing factor to road accidents; however, detecting early visual indicators of driver tiredness remains challenging under realistic driving conditions. This study introduces an artificial intelligence–based system for distinguishing between alert and fatigued drivers using facial images captured in natural driving environments. A total of 41,793 annotated facial images from the Driver Drowsiness Dataset (DDD) were used in the experiments. Although the dataset reflects realistic driving scenarios captured by dashboard-mounted cameras, the proposed system was evaluated offline and not tested in live traffic environments. Deep visual features were extracted using the SqueezeNet architecture and subsequently classified using three supervised learning models: Artificial Neural Networks (ANN), Random Forests (RF), and Support Vector Machines (SVM). Among the evaluated classifiers, ANN achieved the highest performance with an accuracy of 99.97%, followed by RF with 99.78% and SVM with 96.33%. The results indicate that combining lightweight deep feature extraction with classical machine learning classifiers can yield highly accurate fatigue detection while maintaining computational efficiency. The proposed framework provides valuable insights into the development of efficient, real-time driver fatigue monitoring systems with potential applications in accident prevention and road safety enhancement.

Generative AI models can now produce artwork that is virtually indistinguishable from human-created art, posing critical verification challenges for art galleries, educational institutions, and digital content platforms. Existing detection approaches face notable limitations: handcrafted feature–based methods offer interpretability but achieve limited accuracy, while deep learning approaches typically require substantial computational resources and provide minimal explanation for their predictions. We propose an attention-guided fusion framework that integrates Discrete Cosine Transform (DCT)–based frequency-domain features with deep learned representations through a learned attention mechanism, enabling both improved detection performance and interpretable decision-making grounded in signal processing theory. The attention module dynamically weights each feature modality based on input-specific reliability patterns, allowing adaptive fusion across diverse artistic styles and generation methods. We evaluate the proposed framework using two convolutional backbones: the lightweight MobileNetV2 for efficiency-critical deployment scenarios and the higher-capacity ResNet50 for settings where computational resources permit stronger feature extraction. Experiments are conducted on 18,288 artwork images spanning traditional paintings, digital illustrations, and outputs from multiple generative models, using stratified train–validation–test splits and five random seeds for statistical robustness. Under frozen-backbone settings, MobileNetV2-based attention fusion achieves an F1-score of 90.1%, while ResNet50-based attention fusion reaches 90.5%. When backbones are fine-tuned end-to-end, incorporating handcrafted features via attention fusion yields consistent performance gains: MobileNetV2 improves from 93.9% to 94.5% F1-score (+0.6%), and ResNet50 improves from 94.2% to 95.1% F1-score (+0.9%). Feature importance analysis further reveals that low-frequency DCT energy is the most discriminative handcrafted feature, confirming that frequency-domain characteristics effectively distinguish AI-generated from human-created artwork across diverse artistic styles. These results demonstrate that attention-guided fusion of signal processing features and deep learned representations provides consistent accuracy–efficiency benefits across both lightweight and heavyweight architectures, offering a practical and interpretable solution to the emerging challenge of AI-generated artwork detection.

Retrieval-Augmented Generation (RAG) enhances the capabilities of Large Language Models (LLMs) by integrating external knowledge retrieval into the generation pipeline, enabling responses that are grounded, adaptive, and up to date. While RAG can improve factual accuracy compared to models relying solely on pre-trained data, its effectiveness in practice depends strongly on the quality, relevance, and interpretability of retrieved context, and does not eliminate hallucinations entirely. Recent architectures such as Fusion-in-Decoder, Atlas, and ColBERT-RAG demonstrate measurable gains in retrieval precision, scalability, and cross-domain generalization. However, persistent challenges remain, including retrieval noise that can override model reasoning, hallucinations that persist even with high-quality evidence, latency constraints that hinder real-time deployment, and fragile domain adaptation. Moreover, although partial metrics and task-specific benchmarks exist, the absence of a unified evaluation framework for retrieval–generation grounding and robustness complicates fair comparison and reproducible progress. Rather than offering an exhaustive survey, this review provides a focused analytical perspective on retrieval design and architectural evolution in RAG systems. It consolidates representative architectures while critically examining structural limitations related to retrieval–generation coupling as a design choice, context over-reliance, and privacy-preserving computation. Building on these insights, the paper outlines future research directions, including structured knowledge integration via GraphRAG, modular agent-based orchestration in Agentic RAG, improved retrieval filtering, and unified evaluation methodologies. As RAG architectures continue to evolve rapidly in a pre-standardization phase, a more analytically grounded understanding of their design trade-offs is essential for advancing trustworthy and domain-adaptive language systems

This review examines the rapidly expanding landscape of Generative Artificial Intelligence (GenAI) in healthcare, focusing on how models such as GANs, VAEs, diffusion models, and large language models are being explored across medical imaging, clinical documentation, synthetic data generation, drug discovery, and decision-support workflows. Despite GenAI's growing influence, persistent challenges, including limited annotated datasets, concerns over model generalizability, privacy risks, and the opacity of generative architectures, underscore the need for careful evaluation and governance. Accordingly, this study aims to map current applications, assess methodological and ethical constraints, and identify future research opportunities. Using a structured search across ScienceDirect, Scopus, and other sources, the study follows a structured narrative review complemented by quantitative descriptive analysis of the literature. The review also adopts PRISMA-guided screening and standardized data extraction, the review synthesizes evidence from 110 studies published up to October 2025. The findings indicate that the literature frequently reports improvements in imaging quality, data augmentation, molecular modeling workflows, and clinical documentation through generative approaches, particularly in technically constrained settings; however, evidence of clinically validated impact remains uneven across domains. While issues of bias, hallucination, and limited interpretability persist as significant obstacles, imaging-focused applications appear comparatively more mature than decision-support and patient-level modeling tasks. Across domains, diffusion models are commonly associated with higher visual fidelity in biomedical image generation, whereas LLMs demonstrate promise in narrative-oriented tasks but require stronger factual grounding and external verification mechanisms. Overall, the evidence suggests that GenAI’s potential in healthcare is highly context-dependent and contingent upon robust validation frameworks, transparent governance, and human-in-the-loop oversight. The review concludes that responsible integration of GenAI—guided by ethical, legal, and clinical safeguards, will be essential for ensuring safe, equitable, and sustainable adoption in healthcare research, delivery, and policy.

Face detection in low-light conditions remains challenging due to underexposure, noise, and unstable contrast, which significantly degrade the performance of convolutional-based models. Conventional enhancement techniques, such as Histogram Equalization, Contrast-Limited Adaptive Histogram Equalization (CLAHE), and Low-Light Image Enhancement (LIME), often improve brightness but introduce visual artifacts and lack robustness for face detection. This study proposes Low-light Unified Multi-stage Enhancement Network (LUMEN), a lightweight multi-stage image enhancement pipeline designed as a preprocessing module to improve RetinaFace-based face detection under low-light conditions. LUMEN integrates Multiscale Retinex with Color Restoration, adaptive gamma correction, CLAHE-based local contrast enhancement, controlled image fusion, and Non-Local Means denoising to jointly stabilize illumination, preserve texture, and maintain visual naturalness. The method is evaluated on a low-light subset of the Human Faces Object Detection Dataset using RetinaFace. Detection performance is assessed using detection rate and confidence score, while visual quality is evaluated using no-reference metrics such as the Natural Image Quality Evaluator (NIQE) and the Blind/Referenceless Image Spatial Quality Evaluator (BRISQUE). Experimental results show that LUMEN achieves a face detection rate of 91% with a high confidence score (0.9545), rep-resenting an improvement of 33 percentage points over raw low-light images (58%), while maintaining detection performance comparable to CLAHE, which ranks as the second-best method, LUMEN delivers superior perceptual quality, evidenced by the lowest BRISQUE score, as well as more stable visual appearance with reduced noise amplification and fewer contrast artifacts in low-light facial regions. LUMEN also achieves the lowest BRISQUE score, indicating superior texture preservation and perceptual stability. Ablation studies confirm that Retinex and CLAHE are the most critical components for detection robustness, while gamma correction, fusion, and denoising mainly contribute to visual naturalness. These results demonstrate that LUMEN provides an effective and practical preprocessing solution for low-light face detection.

This study investigates how Artificial Intelligence (AI)-based Tutoring Systems can be used to provide secondary school students with personalized learning feedback and to enhance their academic performance. The two main empirical objectives were to assess individualized formative assessment in classroom settings and to examine teachers’ and students’ perceptions of AI adoption. A mixed-methods quasi-experimental design was employed, integrating quantitative pre-/post-tests, interviews, and focus groups across six schools (n = 600). This approach enabled triangulation between measurable learning outcomes and contextual perception data for robust validation. Quantitative data were analyzed using Python (t-tests, ANCOVA), while thematic coding in NVivo was applied to qualitative data. Expert review, pilot testing, and Cronbach’s α (>0.80) were used to validate the instruments and ensure reliability, including pre-/post-tests and engagement scales. Findings revealed that students who received AI-based interventions achieved significantly higher academic performance (Cohen’s d = 1.05) and engagement (d = 0.72) compared with control groups. Teachers with AI exposure reported greater preparedness (mean = 3.4) and fewer perceived barriers. The study provides empirical evidence on the pedagogical viability of AI tutoring in under-resourced contexts, contributing to self-regulated and socio-technical learning theories. It also recommends enhanced systemic teacher education, ethical leadership, and structural support to foster equitable adoption of AI in education. The findings carry strong implications for policy development and educational innovation in promoting data-driven, inclusive learning within Nigeria’s secondary education system

Sentiment analysis is a critical task in natural language processing (NLP) that classifies text into sentiment categories, such as positive, negative, or neutral. This task is particularly challenging for languages like Persian due to the complexity of their linguistic structure and the scarcity of high-quality labeled datasets. Previous studies on Persian sentiment analysis have largely relied on TF-IDF representations or deep learning models, often overlooking handcrafted statistical and stylistic features that capture subtle textual patterns. This limitation reduces their effectiveness, especially when dealing with informal or noisy text data. Experiments in this study were conducted on a dataset of Persian product reviews from Digikala.com, labeled according to user ratings to indicate positive, negative, or neutral sentiment. In this paper, we propose a novel approach to Persian text sentiment analysis by combining statistical and stylistic (surface-level) features with traditional text-based features such as Term Frequency–Inverse Document Frequency (TF-IDF). Unlike prior works that rely solely on TF-IDF or deep learning representations, our method integrates stylistic and statistical cues to capture expressive nuances in informal Persian text. Additionally, the Support Vector Machine (SVM) classifier is optimized using RandomizedSearchCV to enhance performance. The proposed system utilizes both statistical and textual features to improve classification accuracy. We compare its performance with four baseline models, i.e., Naïve Bayes, Logistic Regression, Random Forest, and Decision Tree, that rely solely on TF-IDF features. The experimental results demonstrate that the proposed approach outperforms the baseline models in terms of accuracy, F1-score, recall, and precision. Specifically, the proposed system achieved the highest accuracy (0.8354), significantly improving negative sentiment detection while maintaining strong performance in positive sentiment classification.

Bone fractures are among the most common global health problems and require rapid and accurate diagnosis to prevent complications and reduce healthcare costs. Traditional diagnostic methods rely on the manual examination of X-ray images by experienced radiologists, a process that is time-consuming and prone to human error, especially in emergency departments with high patient volumes. This study proposes a lightweight hybrid approach that integrates deep learning–based feature extraction with classical machine learning algorithms for automatic bone fracture detection. The dataset used comprises 4,840 fractured and 4,623 non-fractured bone X-ray images obtained from the Kaggle platform, in which data augmentation was previously applied by the dataset developer. A compact convolutional neural network architecture, MobileNetV2, was employed exclusively for deep feature extraction. The extracted feature vectors were then used to train and evaluate five ensemble learning algorithms: Random Forest, Extra Trees, Gradient Boosting, AdaBoost, and Bagging Classifier. Model performance was assessed using 10-fold cross-validation based on accuracy, precision, recall (sensitivity), and F-score metrics. Experimental results show that the Extra Trees algorithm achieved the highest performance with 99.14% accuracy, 99.22% precision, 99.03% recall, and 99.12% F-score, while also yielding the shortest training time of 199 seconds. Random Forest followed with an accuracy of 98.38%. The superior performance and computational efficiency of the Extra Trees algorithm demonstrate its suitability for real-time clinical applications. These findings confirm that combining deep feature extraction with ensemble learning can effectively support automated diagnostic systems for bone fracture detection, enabling healthcare professionals to make faster and more reliable decisions.

We introduce DiFACE2 causal DAG, a framework for generating Diverse, Feasible, and Actionable Counterfactual Explanations under causal Directed Acyclic Graph (DAG) constraints. Leveraging background knowledge from an educational intervention program in post-conflict Nigeria, we construct an application-specific causal DAG to identify causal relationships and immutable features during counterfactual generation for black-box ML predictions. Using this DAG, we develop a causally constrained sparse ML model that integrates prior frameworks to generate diverse, feasible, and actionable counterfactuals for the Strengthening Education in Northeast – Early Grade Reading Assessment (SENSE-EGRA) dataset. DiFACE2 produces up to four counterfactuals per prediction, demonstrating high feasibility and actionability for categorical variables. While numeric features occasionally yield infeasible recommendations, this is mitigated by additional non-causal constraints. Although the sparse ML model yields lower accuracy and F1 scores than benchmark frameworks, this stems from the higher complexity and noisiness of the real-world dataset rather than methodological weakness. This trade-off underscores that meaningful, interpretable counterfactuals are preferable to near-perfect accuracy on simplified benchmarks. Therefore, when fairly evaluated on the same complex dataset, DiFACE2 outperforms prior methods across all counterfactual quality metrics.

The pursuit of Artificial General Intelligence (AGI) requires learning frameworks that not only optimize task performance but also align with complex human values. Reinforcement Learning with Human Feedback (RLHF) has emerged as a promising approach to address this challenge; however, conventional RLHF pipelines face scalability issues, reward-model brittleness, and safety concerns. In this study, we propose a PPO-based RLHF framework enhanced with hybrid human–AI oversight and predictive reward evaluation metrics. The framework integrates human annotations with AI-generated critiques, improving data efficiency and robustness against reward hacking. Experimental evaluations on language alignment and control benchmarks demonstrate that the proposed approach achieves a preference win-rate of 78% (vs. 65% in standard RLHF and 54% in supervised fine-tuning), improves task accuracy to 83% (a 12% increase over Sparrow), and reduces safety violations by 31% compared to baseline RLHF. Furthermore, the hybrid oversight strategy enhanced sample efficiency by 1.5×, reducing overall training episodes and annotation costs. These results confirm that the proposed method significantly improves alignment, efficiency, and safety, positioning RLHF with hybrid oversight and predictive evaluation as a practical substrate for advancing safe and scalable AGI systems.

Earthquake damage prediction is essential to improve preparedness and resilience in seismically active regions such as Nepal. This study investigates the application of machine learning models to predict the grade of building damage during seismic events, using building attributes from the 2015 Gorkha earthquake dataset. We employed various algorithms, including Random Forest with both Random Oversampling and SMOTE, Multilayer Perceptron (MLP), Support Vector Machine (SVM), Gradient Boosting, and XGBoost, to tackle class imbalance in damage grades and enhance predictive accuracy. Key building attributes such as construction materials, number of floors, and building age were analyzed to forecast the severity of the damage, which was categorized into three grades. Our findings indicate that the Random Forest model, particularly with SMOTE oversampling, outperformed other techniques, achieving a macro F1-score of 0.67. Both Random Oversampling and SMOTE significantly improved performance over the imbalanced baseline, with Random Oversampling achieving a macro F1-score of 0.65. However, all models faced challenges in accurately predicting Grade 2 damage, underscoring the need for further refinement to address intermediate damage levels. The study highlights the strengths and limitations of various machine learning approaches and emphasizes the significance of feature selection and data balancing methods in improving earthquake damage prediction. It offers valuable insights for policymakers and engineers aiming to enhance earthquake preparedness and develop resilient infrastructure.

This paper examines current trends in Artificial Intelligence (AI), with a particular emphasis on the application of the Python programming language in educational settings. The existing literature reveals a gap in reviews regarding Python's role in AI applications within the education sector. This paper highlights and describes the most effective and popular Python libraries, particularly those relevant to AI in educational settings, that have emerged in recent years. Additionally, it presents basic and widely used AI applications in education that utilize Python. The research method employed is a state-of-the-art review, which involves searching the literature using specific keywords and applying inclusion and exclusion criteria for quality and reliability. In the findings, synthesis, and discussion section, results are presented in tables to facilitate use by researchers and practitioners, allowing for easier application. The tables include information on the most well-known and frequently used Python libraries and AI applications, the grouping categories of Python libraries, their tasks and functions, as well as how Python libraries can be applied in education. Moreover, the paper explores AI methods that enhance teaching and learning using Python tools and connects learning theories with these AI methods. During the discussion, the paper provides an analysis to assist new researchers and practitioners, especially those working with AI in education, on how to effectively leverage AI through Python. It also emphasizes the overall benefits that Python brings to the field of AI in education. The paper also offers various suggestions for addressing gaps in the existing literature.

Depression remains one of the most serious mental health challenges worldwide and is often underdiagnosed due to social stigma and limited access to medical services. With the proliferation of social media as a medium for self-expression, these platforms provide new opportunities for early detection of depressive symptoms through digital footprints. However, most prior research has primarily focused on linguistic features derived from text, overlooking behavioral dynamics that also reflect psychological states. To address this gap, we propose Depress-HybridNet, a hybrid deep learning framework that integrates linguistic embeddings with behavioral activity patterns. The architecture combines a BERT-BiLSTM encoder for linguistic feature extraction, a multi-layer perceptron for behavioral feature encoding, and an adaptive attention-based fusion mechanism to integrate multimodal signals optimally. Experiments conducted on the publicly available Kaggle Depression Dataset demonstrate that Depress-HybridNet consistently outperforms strong baselines, including fine-tuned BERT, achieving state-of-the-art results (F1 = 0.92, AUC = 0.93). A further ablation study highlights the critical role of behavioral features and the attention fusion layer in improving performance. In addition, clinical validation by licensed psychologists confirmed a high degree of consistency between the model’s predictions and expert judgment, underscoring its real-world applicability. These findings underscore the importance of modelling depression as a multifaceted phenomenon, rather than a purely linguistic task, and establish a reproducible benchmark for future research.

Accurate classification of dry bean varieties is essential for enhancing agricultural productivity, ensuring seed quality, and supporting market standardization. Traditional classification methods are time-consuming, labor-intensive, and prone to error, particularly when dealing with morphologically similar varieties. While recent data-driven approaches have shown promising performances, they often suffer from limited varietal representation, inadequate handling of high-resolution image data, and suboptimal dimensionality reduction techniques. To address these challenges, this study proposes a novel hybrid classification pipeline that integrates a fine-tuned InceptionV3 for low- and high-level feature extraction, Principal Component Analysis (PCA) for efficient dimensionality reduction, and a Support Vector Machine (SVM) for final classification. The model was evaluated on a comprehensive dataset combining 14 dry bean varieties from Dogan et al. with five locally sourced varieties, totaling 19 distinct classes. Extensive experimental results showed that the proposed model achieved an accuracy of 94.00% and 80.01% on benchmark and combined datasets, respectively, outperforming state-of-the-art approaches.

Face recognition (FR) is a widely used biometric technology. Nevertheless, achieving efficient and robust FR is still challenging due to variations in illumination, pose, and facial expression. A vital step in any FR system is to select the most informative features and eliminate the redundant ones. In this study, a hybrid approach combining Principal Component Analysis (PCA) and the Dolphin Swarm Algorithm (DSA) with Euclidean Distance as a lightweight classifier is proposed. Experiments were made by using the ORL dataset, which consists of 400 grayscale images. With a 98% recognition rate for this hybrid approach against a recognition rate of 90–92% that can be achieved by PCA only, the proposed PCA+DSA outperformed standalone PCA while still being computationally economical. The metrics of Recognition Rate, Receiver Operating Characteristic (ROC), Cumulative Match Curve (CMC), and Expected Performance Curve (EPC) provided numerous confirmations for this Hybrid model. Additionally, the convergence analysis corroborated DSA’s efficacy in feature selection as the fitness was nearly 92% after nine iterations. Without requiring sophisticated classifiers or deep learning models, our findings show that the identification rate can be improved by combining a bio-inspired optimization technique and the classical PCA method. However, the current study is limited to the ORL dataset in a controlled environment. Future research will focus on implementing and evaluating the system in real-time scenarios on larger and more diverse datasets to enhance its scalability and robustness in practical applications.

Lithium-ion battery demand continues to rise as consumer electronics, hybrid and electric vehicles, and other technologies evolve. This means that numerous lithium-ion batteries are likely to be disposed of, leading to serious disposal problems and negative impacts on the environment and energy conservation. The commonly used lithium-ion battery recycling methods, which are chemical and mechanical, pose challenges, such as some batteries exploding, thermal runaway, or fire. Classification of used lithium-ion battery waste is required to be efficient and reliable. The purpose of this study was to develop a deep learning-based classification model of lithium-ion battery components for automated recycling. A dataset containing images of end-of-life lithium-ion battery components was collected from selected recycling centres in Nairobi, Kenya, and Kaggle.com. The images from selected fields were annotated using the Labelimg annotation tool. The dataset was split into three sets: 70% as the training set, 15% as the validation set, and 15% testing set. A Yolov8n model was then trained using the training set to detect and classify end-of-life lithium-ion battery components. The performance of the model was evaluated using the validation set and test set. The final trained model attained 0.903 precision, 0.792 recall, 0.852 mAP@0.5, 0.724 mAP@0.5-0.95, and 0.844 f1-score. The results from this research study could pave the way for innovative battery recycling measures, guaranteeing that valuable resources are reclaimed and toxic battery waste is managed efficiently and safely.

Brain tumor detection using deep learning has become a focus of research due to its potential to improve patient diagnosis and management. This study presents a comparative analysis to evaluate how the fundamental design philosophies of convolutional neural network (CNN) architectures, namely sequential (VGG16), residual (ResNet50), and encoder-decoder (U-Net), influence performance on brain tumor classification tasks using a public MRI dataset. The data undergoes preprocessing, including normalization, data augmentation, and division into training and testing subsets. Evaluation is conducted using accuracy, precision, recall, and F1-Score metrics. The results show that ResNet50 consistently outperforms other architectures with an average accuracy of 95.35%, followed by VGG16 (93.93%). Conversely, U-Net, designed for segmentation, demonstrates the lowest performance (79%). The superiority of ResNet50 highlights the benefits of residual connections in addressing the vanishing gradient problem for classification tasks. The poor performance of U-Net confirms the hypothesis that architectural suitability for a specific task is more critical than complexity alone. This study highlights the strength of ResNet50 as a reliable approach for brain tumor classification. It underscores the importance of selecting an architecture specifically designed for the intended medical image analysis task. In addition, an analysis of model complexity and computational efficiency was conducted, showing that while ResNet50 achieved the highest accuracy, VGG16 provided a more favorable trade-off between performance and training time.

Credit card fraud detection (CCFD) remains a critical research domain due to the dynamic, adversarial, and highly imbalanced nature of fraudulent activities in financial systems. This study employs a systematic mapping review guided by the PRISMA 2020 guidelines. It analytically synthesizes 40 peer-reviewed and open-access studies, focusing on methodological trends, machine learning techniques, datasets, optimization strategies, and evaluation metrics. Supervised learning (SL) models, including Random Forest, Decision Trees, Support Vector Machine (SVM), and XGBoost, accounted for nearly half of the reviewed studies and consistently demonstrated strong performance. Deep learning (DL) frameworks, including Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and their variants, have demonstrated strong capabilities in capturing sequential and high-dimensional patterns of fraud. However, their effectiveness is constrained by class imbalance and dataset bias. Ensemble and hybrid models further enhanced predictive accuracy but introduced higher computational costs and lower interpretability. A key finding is the heavy reliance on the ECCT 2013 dataset (used in more than half of the reviewed studies), which supports reproducibility but limits generalizability to modern fraud contexts. Optimization strategies, such as the Synthetic Minority Oversampling Technique (SMOTE), hyperparameter tuning, and dimensionality reduction, have proven effective in improving recall and reducing false negatives; however, they have been inconsistently applied. Similarly, evaluation metrics were uneven, with accuracy dominating (reported in 75% of studies), while more informative measures such as recall, F1-score, Precision-Recall curves (AUPRC), and Matthews Correlation Coefficient (MCC) received less emphasis despite their relevance to imbalanced data. Overall, while many models achieve high accuracy in controlled environments, their scalability, adaptability, and trustworthiness in real-world deployment remain limited. Future research should prioritize cross-dataset evaluations, standardized metrics, and emerging paradigms such as federated learning, self-supervised approaches, and explainable AI to guide the development of robust and deployable fraud detection systems.

To meet the development requirements for the complex algorithms with a gradually clear mathematical model and technical routes, frequent modifications and updates for the workflows, and long exploratory cycles, this article conducted preliminary research on a new programming idea based on a visual logic structure - PING programming language. At this time, PING language is a proof-of-concept and is compatible with the C/C++ language. PING language took function modules as the basic programming units. It utilized various explicit execution orders and invoked relationships between function modules to construct the static software logic structure network, implemented a visual scheduling and execution mechanism for function modules, and employed variable symbol table technology to achieve shared variable value transfer. The PING programming language could explicitly represent the overall static logic structure of software during the software programming stage. During the software running stage, the function pointer was used to centrally invoke the function modules, which facilitated the dynamic tracking of the function modules' execution workflows and automatically detected redundant logic branches and untested paths within the software. It had great significance in improving the development, debugging, testing, and maintenance efficiency of interpretable intelligent software, and even in developing machine autonomous programming systems.

This study presents the development of a predictive nutritional assessment system using an Artificial Neural Network (ANN) tailored for adult vegetarians. The system integrates biometric, dietary, and lifestyle data derived from the Anthropometric, Biochemical, Clinical, and Dietary (ABCD) nutritional assessment framework to classify individuals into undernourished, normal, and overweight categories. A feedforward ANN architecture was implemented with 20 input nodes representing ABCD indicators, one hidden layer of 10 neurons, and an output layer of 3 nodes. The model was trained using backpropagation over 1000 epochs with a learning rate of 0.5 and momentum of 0.9, employing data normalization and one-hot encoding for categorical variables. The dataset was split into 80% for training and 20% for testing. The proposed system achieved a classification accuracy of 92.4%, with a 65.3% reduction in cross-entropy loss and an average precision of 89.7% across classes, while processing each input in under 0.8 seconds. Nutritional status predictions were validated using BMI thresholds and cross-referenced with clinical and dietary indicators. Compared to traditional assessment methods, the ANN significantly reduces processing time, improves scalability, and enables dynamic feedback through evolving health indicators. Features such as structured meal plans, real-time daily nutrition summaries, and tailored exercise recommendations enhanced user engagement and adherence to dietary guidelines. The results demonstrate the ANN’s potential as a reliable and scalable tool for personalized nutritional evaluation, enabling early intervention and adaptive planning to address the specific dietary needs and nutrient deficiencies commonly found in vegetarian populations, ultimately supporting predictive healthcare and dietary optimization.

This study proposes a lightweight hybrid model that integrates probabilistic feature augmentation, Gated Recurrent Units (GRU), and Multi-Head Attention to enhance chronic disease prediction on imbalanced clinical tabular data. The research addresses the challenge of low recall and poor minority-class detection in conventional models, aiming to improve predictive robustness and interpretability. The proposed model leverages logistic regression to generate probability-based feature augmentations, which are combined with sequential dependencies learned by GRU and refined through attention mechanisms. Evaluated on three benchmark medical datasets, Breast Cancer, Heart Disease, and Hepatitis C, the model achieves a maximum F1-score of 0.951, a recall of 0.944, and an AUC of 0.976, outperforming traditional machine learning baselines and single-path deep learning models. The attention module enhances interpretability by highlighting relevant features, supporting clinical insights. These findings confirm that probabilistic augmentation and attention-guided deep architectures can significantly improve prediction performance on imbalanced medical data. The results support the study’s objective to design an accurate, interpretable, and clinically relevant prediction model.

Indoor positioning technology plays an important role in improving efficiency and automating industrial processes such as warehouse management, production lines, and mobile robot navigation. However, existing RFID-based and odometry-only localization methods still suffer from limited accuracy, drift, and dependence on predefined infrastructure. To address these challenges, this work proposes a lightweight sensor fusion framework that combines wheel encoder data and phase-based RFID signals using an Extended Kalman Filter (EKF) for accurate indoor localization of mobile robots. The proposed method does not require prior knowledge of the tag map and enables convergence even when the robot starts outside the reader's range. Simulation results demonstrate that the fused method achieves an average positioning error of 5.4 cm and a final error of less than 8 cm. An ablation study comparing odometry-only, RFID-only, and fusion scenarios confirms the superiority of the integrated approach in terms of accuracy and robustness. The system is suitable for real-time implementation in cost-effective embedded platforms and has potential for deployment in smart warehouses and logistics environments.

Agricultural yields have significantly decreased over the years due to soil degradation and weather variability. Inefficient utilization of these farming resources has worsened the situation. To address this challenge, this study developed an AI-driven crop recommendation framework based on structured soil and climate data. The dataset comprises key features such as temperature, humidity, and rainfall, as well as soil pH, nitrogen, phosphorus, and potassium, collected from the Wannune axis of Tarka Local Government Area in Benue State, Nigeria. To handle the data imbalance in the dataset, the Synthetic Minority Over-Sampling Technique (SMOTE) was applied. Recursive Feature Elimination (RFE) was employed to optimize feature selection. Five tree-based models, including Random Forest, Decision Tree, Naïve Bayes, Logistic Regression, and Extreme Gradient Boost (XGBoost), were evaluated. The XGBoost model yielded the best performance, with an accuracy of 99.65%. The use of indigenous data bridges the gap between reliability and the relevance of localization in precision agriculture. Future work includes integrating dynamic environmental variables and conducting field validation to improve scalability and adoption

Breast cancer remains one of the leading causes of death among women worldwide, primarily due to late detection and diagnosis. Early and accurate prediction is essential to improve survival rates. Machine learning (ML) techniques have proven effective in supporting early diagnosis. This study aims to evaluate and compare the performance of three different approaches: traditional ML, ensemble ML, and deep learning (DL) for early prediction of breast cancer using the Wisconsin Diagnostic Breast Cancer (WDBC) dataset. The methodology includes data collection, preprocessing, and the design of predictive models. Traditional ML algorithms used include Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naïve Bayes (NB), and Decision Tree (DT). Ensemble ML techniques comprise Random Forest (RF), XGBoost, and AdaBoost, while DL models include Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN). The models were evaluated using precision, recall, F1-score, and accuracy. The results indicate that XGBoost achieved the highest accuracy (0.99), with strong recall (0.98) and F1-score (0.986), outperforming all other ensemble and traditional ML methods. CNN achieved 0.99 in all evaluation metrics, slightly outperforming RNN, which attained 0.98 accuracy and 0.985 F1-score. These findings confirm that ensemble ML techniques outperform traditional models, while CNN leads among DL models. Furthermore, the proposed models demonstrated superior prediction performance compared to existing studies, particularly in minimizing false negatives, which is critical for healthcare applications.

This paper presents a vehicle detection, tracking, and counting system for urban traffic videos based on integrating YOLOv9s and DeepSORT. The proposed method aims to address challenges such as occlusions, high vehicle density, and identity consistency in surveillance video analysis. A private dataset consisting of annotated traffic videos recorded in Thai Nguyen, Vietnam, was used for evaluation. The YOLOv9s model was selected for its balance between speed and accuracy, while DeepSORT provides robust multi-object tracking using appearance features and Kalman filtering. Experimental results demonstrate that the system achieves a mean Average Precision (mAP at 0.5) of 91.4%, an mAP at 0.5-0.95 of 82.7%, and operates at an average speed of 12.9 frames per second (FPS). Vehicle counting was validated against manually annotated ground truth with an average error rate of less than 4%. These results indicate that the proposed approach is both accurate and efficient for real-time traffic monitoring applications.

This paper presents a novel ingredient-based recipe recommendation system that suggests dishes using only available ingredients, eliminating reliance on user preference data. As online culinary platforms host millions of recipes, users struggle to find dishes that match their pantry constraints, dietary needs, and culinary preferences. Unlike traditional methods like TF-IDF, which lack semantic depth, or existing methods such as graph neural networks (GNNs), which incur high computational costs and struggle with cold-start scenarios, our self-supervised contrastive learning approach maps ingredients and recipes into a 64-dimensional semantic space, capturing complex culinary relationships for precise, scalable recommendations. The system enhances representation quality without user history by leveraging dual-modal Word2Vec embeddings that incorporate ingredients and cooking instructions and optimizing them using triplet loss. Using a dataset of 122,265 standardized recipes, our model achieves a 79.11% ingredient match ratio, 70% recommendation diversity (vs. 50% for baselines), and sub-100ms query times. It outperforms TF-IDF and KNN by 14% in relevance. The implementation addresses ingredient normalization, rare ingredient handling, and diverse cuisine representation, promoting sustainable cooking by maximizing ingredient use and reducing food waste. This work advances the methodology of recommendation systems by addressing the cold-start problem and ensuring privacy through user-data-free modeling. It supports practical culinary applications, contributing to sustainable cooking practices and culinary informatics.

Object detection for sensing images is one of the promising research directions in computer vision. Applications for object detection from remote sensing images play an important role in analyzing aerial or satellite imagery. Benefits include applications in monitoring buildings and infrastructure, transportation, supporting search and rescue or responding to natural disasters, and environmental research. However, detecting objects in remote sensing images is difficult due to the diversity of shapes and sizes, viewing angles of objects, and complex background environments. In this paper, the authors present a Deep Learning (DL)-based object detection process from remotely sensed images, the main goal of which is to improve the ability to detect small objects in high-resolution aerial images. Implement and evaluate the super-slicing inference technique in the YOLOv11 model to improve the ability to detect very small and extremely small objects. Many simulation results are tested experimentally in the problem of detecting and tracking vehicles in Vietnam (Thai Nguyen). The results show that the system can accurately detect small objects such as pedestrians, motorbikes, and cars at a distance, with confidence ranging from 0.31 to 0.90. Some detection situations are successful even when the object is located at the edge of the slice. Finally, the authors discuss potential future research directions and unaddressed formulations.

Hyperspectral unmixing is a fundamental task in remote sensing that aims to decompose mixed spectral pixels into their constituent materials and estimate their respective abundances. Traditional methods often face limitations due to nonlinear mixing effects and the lack of pure pixels. This paper proposes a deep learning-based comparative framework that integrates modified versions of the Minimum Simplex Convolutional Network (MiSiCNet) and Unsupervised Deep Image Prior (UnDIP), referred to as Blind MiSiCNet and Supervised UnDIP, respectively, to achieve robust unmixing and abundance estimation in real-world scenarios. The proposed Blind MiSiCNet removes downsampling and upsampling layers and leverages spatial and geometric priors through convolutional layers with a minimum simplex volume constraint, producing reliable initial abundance maps even in the absence of pure pixels. The Supervised UnDIP variant further refines these estimates using the implicit regularization of convolutional neural networks, incorporating known endmembers to generate noise-free and spatially coherent abundance maps. Experimental results on the real-world Jasper Ridge dataset demonstrate that the proposed supervised and blind unmixing methods, evaluated comparatively, significantly outperform existing approaches regarding Spectral Angle Distance (SAD) and Root Mean Square Error (RMSE). The results also highlight improved noise robustness and better preservation of spatial structures.

This study presents RetenNet, a comprehensive framework for managing customer churn in telecommunications, integrating predictive modelling, prescriptive optimization, and explainable artificial intelligence (XAI) incorporated with Large Language Models (LLMs). The process commences with the IBM Telco dataset, divided in an 80:20 ratio into training and testing sets. Categorical variables are converted by one-hot and label encoding, whilst class imbalance is mitigated using SMOTEENN. Min-max scaling and mutual information-based feature selection guarantee data appropriateness for machine learning models. Five classification algorithms, i.e., Random Forest (RF), Support Vector Machine (SVM), Extreme Gradient Boosting (XGB), Logistic Regression (LR), and Multi-Layer Perceptron (MLP) are assessed. The SVM model utilizing an RBF kernel exhibits optimal performance. In conjunction with nested cross-validation, Bayesian optimization guarantees excellent hyperparameter optimization and generalization. Performance is evaluated using the F1-score to highlight the implications of false negatives and false positives in churn situations. The methodology additionally incorporates fuzzy rule-based clustering, facilitating flexibility in customer segment identification for intervention priority. Prescriptive optimization uses linear integer programming to distribute retention budget according to model results and business constraints. SHAP waterfall plot employed to guarantee transparency and facilitate actionable insights. Furthermore, Gemini 1.5 Flash, a multimodal LLM, generates analysis and produces contextual recommendations derived from the SHAP waterfall plot. RetenNet offers a comprehensive and interpretable approach to the churn management pipeline, including classical machine learning, prescriptive optimization, and LLM-driven explainable artificial intelligence to enhance decision-making in customer retention efforts.

In recent times, the control of engineering processes has gained attention, making it critical to protect devices from operating beyond their designed capabilities. Therefore, this study investigates the modelling and optimization of gain factors for a MIMO-based PID controller employing a machine learning classifier approach. Traditional approaches like Ziegler-Nichols and Cohen-Coon may not handle dynamic coupling between channels well, struggle with nonlinearity, and always require manual tuning, which is tediously challenging and time-consuming. Machine-learning (ML) algorithms are introduced in this study to redesign and remodel the general model (GM) in MIMO fashion. We train and test the proposed model in Anaconda3 using Python programming on synthetic and iteratively generated datasets. Four different ML classifiers, namely, Support Vector Machine (SVM), Random Forest (RF), Logistic Regression (LR), and Naïve Bayes (NB), are employed to generate confusion matrices with accuracies of 30.67%, 72%, 25.33%, and 76%, respectively, using the PM system. Furthermore, the dataset corresponding to 76% of the NB is predicted to give an optimal output, as the output of the GM serves as the initial solution of the controller, and the test numerical results are fed back into the MATLAB scripts to produce an optimized version of the GM. Comparatively speaking, the GM, with an average settling time of 238.33 s, and the PM, with an average of 93.33 s, demonstrate faster steady-state response times, smaller gain, phase margins, and better overlap of the ideal gain factors on their target open-loop gain curves, depicting excellent stability. In addition, it can be inferred that the PM is more efficient than the existing model (GM) as it yields better-desired results. As a future direction, it is suggested that more machine learning classifiers, such as XGBOOST, neural networks, and K-nearest neighbour, be incorporated into the training and testing of larger datasets for better, improved, and optimized outputs.

Distributed Denial of Service (DDoS) attacks remain a critical threat to network infrastructure, necessitating timely and precise detection techniques to mitigate their impact. This study presents a hybrid deep learning framework that integrates the Three Conditions for Feature Aggregation (3ConFA) framework with a one-dimensional Convolutional Neural Network (1D-CNN) for effective DDoS detection. Initially, salient features were selected using the 3ConFA approach, which combines multi-filter feature ranking—based on Chi-square, Information Gain, and Decision Tree Recursive Feature Elimination (DT-RFE) to extract robust and relevant features from raw network traffic data. The training samples were balanced using the Adaptive Synthetic Sampling Approach (ADASYN) to address the high-class imbalance typically present in DDoS datasets. The refined features were then passed through a 1D-CNN model designed to learn temporal and spatial attack behavior patterns. Feature fusion was applied by concatenating the aggregated statistical features and the deep features learned by the CNN, then re-selected the most informative features using Recursive Feature Elimination with Cross-Validation (RFECV). The final classification was performed using a Softmax output layer, and the model was evaluated using 5-fold cross-validation and a separate test set. Experimental results demonstrated an average training accuracy of 99.42%, an F1-score of 99.35%, and an AUC-ROC of 99.87%. On the test set, the model achieved a detection accuracy of 99.56%, a precision of 99.61%, and an F1-score of 99.50%, with an AUC-ROC of 0.9982. The proposed hybrid approach outperforms traditional models such as Random Forest, Decision Tree, XGBoost, and standalone CNN, validating the synergistic impact of integrating 3ConFA with deep temporal convolutional modeling for accurate and interpretable DDoS attack detection.

Nowadays, the construction, monitoring, quality control, and maintenance of roads always require high-precision, easy-to-use measuring devices in the field. This study develops an embedded computer program using the YOLOv8 model integrated into a smart road length measuring device to detect road surface cracks. First, the study analyzes and clarifies the outstanding points of the YOLOv8 model, including anchor-free: Eliminates the use of traditional anchor boxes, helping to simplify the training process, increase accuracy and reduce computational costs; C2f (Cross-Stage Partial with feature fusion); Replacing CSSPLayer in YOLOv5, this module improves feature extraction and maintains good computational performance; SPFF (Spatial Pyramid Pooling - Fast): Helps expand the receptive field without increasing computational costs, supporting object recognition at multiple scales; PAN++: Enhances the transfer of features from lower layers to the output, helping to detect small objects such as cracks well. Then, this work evaluated the performance of the proposed model by applying YOLOv8 on the surface crack dataset taken from the open-source Roboflow Universe. The results showed that the developed YOLOv8 model achieved good results with the recall indexes of 57.3%, mAP50 59.3%, precision of 64.4% mAP50-95 48.9% for YOLOv8n model, and can improve the accuracy further, reaching 79.8% using the YOLOv8m model, while still meeting the real-time processing speed of the device. These experimental results using the YOLOv8 model for detecting cracks on DongAnh road, Hanoi, Vietnam, confirm the possibility of applying the integrated road crack detection model on smart road length measuring devices in practice, achieved a precision of 43.3%, supporting the survey, inspection, supervision, construction and management of traffic infrastructure more scientifically and effectively

Communication is an essential part of day-to-day life. However, people who have disabling hearing loss, which in general is known as deafness, or have a severe loss of hearing use sign language for their communication. Those people learn this language through their respective schools, whereas hearing people rarely get to know this communication tool, leaving a wide communication gap. To bridge this gap, an automated communication system that can convert text to sign language without depending on human interpreters. This literature review overviews sign language and discusses the different technologies used for text-to-sign conversion systems. The paper evaluates computer-based text processing methods, machine translation techniques (rule-based, corpus-based, and neural architectures), and sign synthesis approaches (avatar animation, video concatenation, neural generation). The paper compares each method's strengths and limitations and presents quantitative performance data from recent systems. The review also provides a comprehensive summary of widely used datasets and introduces a visual taxonomy of translation approaches. Furthermore, the paper discusses the current limitations, including scalability, data limitations, the need for more expressive synthesis, and future scope for improvement. The expected outcome of this review is to find and gather information on various sign language systems worldwide, identify research gaps, and propose further enhancements in text-to-sign language conversion.

Diabetes is the body’s inability to efficiently break down sugar or secrete enough insulin required to process glucose, which supports normal bodily functions. Diabetes, as a prevalent chronic disorder, has contributed to numerous underlying health challenges among its carriers and is classified by the WHO as the world’s deadliest disease and silent killer. Its non-communicable nature makes early diagnosis difficult, allowing progression through various stages: type I, type II, pre-diabetes, and gestational. This challenge is further compounded by the imbalanced nature of diabetes datasets, which leads to high misclassification, poor generalization, and reduced accuracy. This study predicts diabetes using a bi-directional long short-term memory (BiLSTM) model applied to two datasets: (a) PIMA Indian Diabetes and (b) Iraqi Society Dataset, to evaluate the impact of six known balancing techniques and assess their effectiveness. Results show that for PID, the SMOTE-Tomek fused BiLSTM outperforms other balancing schemes with F1, Accuracy, Precision, Recall, and Specificity scores of 0.9182, 0.9198, 0.9128, 0.9248, and 0.9208, respectively. For ISD, it also achieves the best performance with values of 0.9367, 0.9369, 0.9386, 0.9388, and 0.9313, respectively. Other balancing approaches yielded F1 scores ranging from [0.6751 to 0.9347], accuracy [0.684 to 0.9358], Precision [0.6851 to 0.9296], Recall [0.6639 to 0.9356], and specificity [0.6658 to 0.9298]. These results imply that BiLSTM is resilient to the vanishing gradient problem and can effectively classify diabetes cases with enhanced performance.

This paper presents recent research in the field of human face recognition and detection. Technology development has marked a clear transformation in the implementation of recognition tasks. It is a transition from classical methods to methods with AI and deep learning applications. We have reviewed the latest works in famous journals to evaluate this field. This paper provides a comprehensive taxonomy across detection, identification, and tracking. It offers a comparative analysis of state-of-the-art AI approaches and highlights their strengths, limitations, and practical considerations. We have evaluated the level of compliance, completeness, and methodology of related studies. According to the results of our evaluation, significant improvements and developments have occurred in this field. However, some challenges still need to be focused on, such as improving the efficiency of training models in different environmental conditions. The negative effects of lighting conditions, occlusion of faces, or distortions due to different shooting angles are overcome.

Indoor positioning plays an important role in a wide range of applications, such as autonomous mobile robots (AMRs), healthcare smart homes, and industrial automation. Popular positioning methods include received signal strength indicator (RSSI)-based positioning, proximity-based positioning, angle-of-arrival (AoA)-based positioning, and fingerprinting. Each method has specific trade-offs regarding accuracy, cost, and deployment complexity. Recent research has focused on solutions to improve accuracy. These include noise filtering techniques, such as Kalman and particle filters, as well as the use of machine learning to optimize the positioning model. In particular, graph-based positioning methods have demonstrated the ability to improve accuracy by optimizing the spatial relationship between landmarks. In addition, regression-based error estimation helps predict and correct positioning errors, improving system reliability. This paper presents a focused review of state-of-the-art techniques to enhance indoor positioning systems' accuracy. They consist of filtering, hybrid methods, and Artificial Intelligence (AI)-based optimization, offering a practical reference for enhancing system performance in real-world applications. By reinforcing recent developments, this review serves as a foundation for researchers and engineers seeking effective positioning solutions for AMR and smart environments. Looking ahead, hybrid positioning systems integrate multiple technologies. AI and machine learning enhance them. Hybrid positioning systems are expected to offer greater accuracy, flexibility, and scalability. These advancements will be essential in optimizing operations and enhancing user experiences in future automation applications and the Internet of Things(IoT).

Secure image transmission is increasingly vital in the digital era, especially against emerging quantum threats. This study proposes a hybrid image encryption scheme that integrates Quantum Key Distribution (QKD) using the BB84 protocol with a combination of 7-dimensional (7D) and 2-dimensional (2D) hyperchaotic systems to achieve robust security. The BB84 protocol facilitates quantum-assisted key exchange, ensuring resistance to eavesdropping, while the hyperchaotic systems provide high entropy and complex randomness, utilized in a layered permutation-substitution encryption framework. The initial seeds for chaotic sequences are derived using a SHA-512 hash of both the input image and quantum-generated key, ensuring uniqueness and sensitivity. Experimental validation was conducted using several benchmark images. The information entropy values of the ciphered images reached up to 7.9993, indicating excellent randomness. Differential analysis showed high resistance to small perturbations, with NPCR exceeding 99.61% and UACI averaging around 33.47%, which meet standard security thresholds. Histogram and chi-square tests confirmed the uniform pixel distribution, with chi-square values below 280, satisfying the randomness criterion for 8-bit images. Furthermore, correlation coefficients of adjacent pixels dropped to near zero, evidencing effective decorrelation. The encryption scheme also demonstrated robustness to data loss, as shown by the successful decryption of partially corrupted cipher images. Robustness testing under partial data loss (200×200-pixel blocks) also demonstrated visual recoverability and algorithm resilience. Overall, the proposed BB84-assisted dual-hyperchaotic encryption scheme offers a secure and computationally effective solution for protecting sensitive image data, making it suitable for post-quantum secure communications.

Indonesia’s waste management system struggles to keep pace with rapid population growth and urbanization, resulting in inefficient waste collection, environmental degradation, and low recycling rates. The country predominantly relies on open dumping and landfilling, which contribute significantly to pollution and greenhouse gas emissions. This study explores the transformative role of Artificial Intelligence (AI) and the Internet of Things (IoT) in waste management, focusing on smart waste collection, automated sorting, real-time monitoring, and predictive analytics. AI-driven waste classification enhances recycling efficiency, while IoT-enabled smart bins optimize collection routes, reducing operational costs and landfill dependency. Despite these advantages, challenges such as high implementation costs, digital infrastructure limitations, and data privacy concerns hinder widespread adoption. This study highlights that policy support, investment in digital infrastructure, and stakeholder collaboration are crucial for successful implementation. By leveraging AI and IoT, Indonesia can significantly improve waste management efficiency, minimize environmental impact, and advance circular economy initiatives. The findings suggest that, with the right policies and investments, AI-driven waste management can drive sustainability, reduce waste mismanagement, and promote resource optimization, making it a vital strategy for future urban development in Indonesia.
 

The rapid evolution of artificial intelligence (AI) has significantly transformed the field of steganography, extending its scope beyond conventional image-based techniques to novel domains such as linguistic and 3D mesh data hiding. This review presents a concise, accessible, and critical examination of recent AI-powered steganography methods, focusing on three distinct modalities: image, linguistic, and 3D mesh. Unlike most surveys focusing solely on one modality, this work highlights some modalities, identifies their unique challenges, and discusses how AI has reshaped embedding mechanisms, evaluation strategies, and security concerns. In image-based steganography, deep models such as GANs and Transformers have improved imperceptibility and extraction accuracy, but face limitations in computational efficiency and extraction consistency. Linguistic steganography, previously hindered by semantic fragility, has been revitalized by large language models (LLMs), enabling context-aware and reversible embedding, though still constrained by metric standardization and synchronization issues. Meanwhile, 3D mesh steganography remains dominated by non-AI methods, offering fertile ground for innovation through geometric deep learning. This review also provides a comparative summary of design principles, performance metrics, and modality-specific trade-offs. The analysis reveals a shift in evaluation paradigms, from numeric fidelity (e.g., PSNR, SSIM) to semantic and perceptual metrics (e.g., LPIPS, BERTScore, Hausdorff Distance). Looking ahead, future directions include cross-modal integration, domain adaptation, lightweight AI models, and the development of unified benchmarks. By presenting recent advances and critical perspectives across underexplored domains, this survey aims to inspire early-stage researchers and practitioners to explore new frontiers of steganography in the AI era.

Expanding extensive and intricate social networks has required sophisticated community detection techniques. This study presents an innovative hybrid methodology that utilizes node space similarity and local knowledge to enhance community identification. Node space similarity is defined by integrating eigenvector centrality (EC), which quantifies a node’s influence inside the network, with proximity metrics, such as closeness, to evaluate the connectivity between nodes. This enables us to identify cohorts of individuals with analogous influence and connectivity. We use local knowledge by concentrating on these pivotal nodes' direct connections and attributes, allowing the technique to broaden community discovery (CD) outward effectively. Our five-phase methodology, grounded in an iterative seed expansion algorithm, commences with identifying highly central nodes and progressively develops communities by integrating nodes exhibiting high similarity and local connectivity. The method incorporates graph statistical inference and embedding features to improve accuracy and capture extensive network patterns. This integrated approach facilitates the precise and effective identification of communities within extensive social networks, exceeding the constraints of conventional techniques. This research attained a modularity of 95.05% on the DBLP dataset and 94.50% on the Amazon dataset. This study achieved a Normalized Mutual Information (NMI) of 91.80% on the DBLP dataset, 92.50% on the Amazon dataset, and 90.43% on the football dataset, demonstrating superior performance relative to previous methodologies. The findings indicate that the hybrid method outperforms other recognized methods in large-scale graphs, showcasing notable robustness and efficiency.

Feature selection in most black-box machine learning algorithms, such as BERT, is based on the cor-relations between features and the target variable rather than causal relationships in the dataset. This makes their predictive power and decisions questionable because of their potential bias. This paper presents novel BERT models that learn from causal variables in a clinical discharge dataset. The causal-directed acyclic Graphs (DAG) identify input variables for patients’ survival rate prediction and decisions. The core idea behind our model lies in the ability of the BERT-based model to learn from the causal DAG semi-synthetic dataset, enabling it to model the underlying causal structure accurately in-stead of the generic spurious correlations devoid of causation. The results from Causal DAG Conditional Independence Test (CIT) validation metrics showed that the conceptual assumptions of the causal DAG were supported, the Pearson correlation coefficient ranges between -1 and 1, the p-value was (>0.05), and the confidence interval of 95% and 25% were satisfied. We further mapped the semi-synthetic dataset that evolved from the Causal DAG to three BERT models. Two metrics, pre-diction accuracy, and AUC score, were used to compare the performance of the BERT models. The accuracy of the BERT models showed that the regular BERT has a performance of 96%, while Clinical-BERT performance was 90%, and Clinical-BERT-Discharge-summary was 92%. On the other hand, the AUC score for BERT was 79%, ClinicalBERT was 77%, while ClinicalBERT-discharge summary was 84%. Our experiments on the synthetic dataset for the patient’s survival rate from the causal DAG datasets demonstrate high predictive performance and explainable input variables for human under-standing to justify prediction.

Artificial intelligence can change many industries as a global phenomenon. Over the years, this transformation has supported Electronic Markets in reengineering the processes and activities that take place in traditional markets, focusing on improving transaction effectiveness and efficiency. While our dependence on intelligent machines continues to grow, the demand for more transparent and interpretable models equally grows. Thus, explanations for machine decisions and predictions are needed to justify their reliability, which requires greater interpretability and often elaborates the need to understand the algorithms' underlying mechanism. This paper, therefore, proposed models based on Decision Tree (DT), Long Short-Term Memory (LSTM), and an ensemble of the two aforementioned models for improving CLV accuracy, interpretability, and explainability of AI-based models in the multi-category electronic market. An open-source e-commerce Behavior Data from a multi-category store, previously used by similar studies on XAI and CLV, was used in this experiment, ensuring the robustness of the product prediction and explanations and fair comparison. From the results, the models from this study demonstrated remarkable performance in terms of minimal error rates of MAE, MSE, and RMSE, with LSTM outperforming the other models. Regarding explainability and interpretation, the begin_time is ranked as the most relevant feature in CLV prediction.

This research presents a hybrid intrusion detection approach that integrates Generative Adversarial Networks (GANs) for synthetic data generation with Random Forest (RF) as the primary classifier. The study aims to improve detection performance in cybersecurity applications by enhancing dataset diversity and addressing challenges in traditional models, particularly in detecting minority attack classes often underrepresented in real-world datasets. The proposed method employs GANs to generate synthetic attack samples that mimic real-world intrusions, which are then combined with real data from the UNSW-NB15 dataset to create a more balanced training set. By leveraging synthetic data augmentation, our approach mitigates issues related to class imbalance and enhances the generalization capability of the classifier. Extensive experiments demonstrate that RF trained on the combined dataset of real and synthetic data achieves superior detection performance compared to models trained exclusively on real data. Specifically, RF trained solely on the original dataset achieves an accuracy of 97.58%, whereas integrating GAN-generated synthetic data improves accuracy to 98.27%. The proposed methodology is further evaluated through comparative analysis against alternative classifiers, including Support Vector Machine (SVM), XGBoost, Gated Recurrent Unit (GRU), and related studies in the field. Our findings indicate that GAN-augmented training significantly enhances detection rates, particularly for rare attack types, while maintaining computational efficiency. Furthermore, RF outperforms other classifiers, including deep learning models, demonstrating its effectiveness as a lightweight yet robust classification method. Integrating GANs with RF offers a scalable and adaptable framework for intrusion detection, ensuring improved resilience against evolving cyber threats.

Phishing is a fraudulent activity wherein an attacker impersonates a trusted individual or organization to acquire sensitive information from an online user. Phishing websites have become a major cyber-security issue in the contemporary digital landscape. As online activities expand in e-commerce, banking, and social media, the hazards presented by these fraudulent websites have intensified. Deep learning-based Natural Language Processing (NLP) approaches offer an effective solution for detecting phishing URLs. However, deploying large models like BERT or RoBERTa for real-time detection poses computational challenges. This study proposes BERTPHIURL, a Teacher-Student Learning framework that leverages RoBERTa as the Teacher model and DistilRoBERTa as the Student model to improve phishing detection efficiency while reducing computational overhead. By applying knowledge distillation, the Student model learns from the Teacher, preserving high detection accuracy with significantly lower resource consumption. The proposed approach effectively captures contextual relevance and local features in malicious URL detection tasks. The experiments were conducted on a dataset exceeding 50,000 URLs to evaluate performance. Results indicate that BERTPHIURL achieves a 94.22% accuracy, outperforming existing phishing detection methods while maintaining efficiency suitable for real-time applications.

The research focuses on developing an Optimized Multiclass Intrusion Classifier (OMIC), an advanced framework for large-scale network intrusion detection in IoT environments. Traditional intrusion detection systems face significant challenges with increasing network complexity, attack sophistication, and the exponential growth of IoT devices, particularly in handling class imbalance, computational efficiency, and real-time processing of massive data volumes. OMIC introduces a novel ensemble approach combining LightGBM and XGBoost classifiers with a memory-optimized processing pipeline to address these limitations. The framework implements sophisticated data handling techniques, including dynamic chunk-based processing, adaptive sampling methods, and cost-sensitive learning to manage class imbalance. Experimental evaluation using the comprehensive CICIoT2023 dataset, comprising over 1 million records and 33 distinct attack types, demonstrates OMIC's exceptional performance with an overall accuracy of 99.26%. The framework achieves perfect precision, recall, and F1-scores for most DDoS and DoS attack categories, significantly outperforming traditional machine learning and deep learning approaches. While excelling in most attack categories, OMIC shows limitations in detecting certain web-based attacks and reconnaissance activities, suggesting areas for future enhancement. The framework's superior performance in handling large-scale data while maintaining high detection accuracy positions it as a significant advancement in IoT network security, offering practical solutions for real-world deployments.

Face spoofing detection is critical for biometric security systems to prevent unauthorized access. This study proposes a deep learning-based approach integrating FaceNet and DenseNet201 to enhance face spoofing detection performance. FaceNet generates identity-based embeddings, ensuring robust facial feature representation, while DenseNet201 extracts complementary texture-based features. These features are fused using the Concatenate function to form a more comprehensive representation for im-proved classification. The proposed method is evaluated on two widely used face spoofing datasets, NUAA Photograph Imposter and LCC-FASD, achieving 100% accuracy on NUAA and 99% on LCC-FASD. Ablation studies reveal that data augmentation does not always enhance performance, particularly on high-complexity datasets such as LCC-FASD, where augmentation increases the False Rejection Rate (FRR). Conversely, DenseNet201 benefits more from augmentation, while the proposed method performs best without augmentation. Comparative analysis with previous studies further confirms the superiority of the proposed approach in reducing error rates, particularly Half Total Error Rate (HTER), False Acceptance Rate (FAR), and FRR. These findings indicate that combining identity-based embeddings and texture-based feature extraction significantly improves spoofing detection and enhances model robustness across different attack scenarios. This study advances biometric security by introducing an efficient feature fusion strategy that strengthens deep learning-based spoof detection. Future research may explore further optimization strategies and evaluate the approach on more diverse datasets to enhance generalization.

This study aimed to predict and detect cyberattacks using hybrid machine-learning models. The CICIoT2023 dataset was utilized for attack prediction and detection, and model performance was evaluated separately by performing thirty-four class (33+1), eight class (7+1), and binary (1+1) classifications according to the types of attacks in the dataset. Voting and stacking hybrid machine learning models were employed in this study, with Logistic Regression (LR), Gaussian Naive Bayes (GNB), and Random Forest (RF) algorithms selected as sub-models. Data preprocessing steps were applied to enhance model performance, and hyperparameter optimization was performed. As a result, this study achieved an accuracy of 98% in thirty-four class classifications, 88% in eight class classifications, and 99% in binary classifications with the Voting hybrid machine learning model. In contrast, the Stacking hybrid machine learning model reached an accuracy of 98% in both thirty-four class and eight class classifications and 99% in binary classifications. This study presents a significant innovation in the cybersecurity field by introducing an innovative approach to eliminating the disadvantages of single-model methods.

Project management is changing drastically due to the integration of artificial intelligence (AI) and quantum computing (QC), redefining traditional methods. This study explores Quantum AI (QAI) and AI-driven solutions to tackle enduring issues, including resource inefficiencies, schedule delays, and budget overruns. These technologies significantly enhance project outcomes by leveraging predictive analytics, dynamic scheduling, and high-dimensional optimization. A comparative analysis of prominent case studies, including the Crossrail Project, East Side Access, and the Montreal Olympics, highlights the superior performance of AI and QAI techniques compared to conventional methods. The study shows that QAI can cut delays by 60%, optimize resource allocation with 83% efficiency, and eliminate cost overruns by up to 40% using Monte Carlo simulations and Failure Mode Effects Analysis. These results demonstrate that quantum artificial intelligence is a ground-breaking tool for handling intricate, interconnected project settings. Additionally, this study emphasizes how QAI is scalable and applicable across industries, especially in fields that need real-time optimization and high-dimensional data processing. The proposed hybrid quantum-classical paradigm provides practical solutions and sets a benchmark for efficiency, scalability, and risk mitigation in project management.

Course recommendation aims to find suitable and attractive courses for students based on their needs, playing a significant role in the curricula-variable system. However, with the abundant available courses, students often face cognitive overload when selecting the most appropriate ones. This research proposes a course recommendation system called the Enhanced Hybrid Course Recommender to address this challenge. This system uses an ensemble learning approach to combine and leverage the power of multiple machine learning classifiers, including Random Forest, Naive Bayes, and Support Vector Machine. By utilizing TF-IDF vectorization for text data transformation and label encoding for target label compatibility, this experiment significantly enhances recommendation precision and relevance, easing students' decision-making process and improving the overall quality of course recommendations. A hybrid approach is applied to improve the recommendation quality by combining predictions from all three classifiers through weighted voting. This ensemble method improves overall robustness and accuracy. This approach not only mitigates the cognitive overload faced by students but also significantly improves the quality of recommendations. Our hybrid model represents a substantial advancement in personalized course recommendation technology by demonstrating superior performance across key evaluation metrics such as accuracy, precision, recall, F1-score, ARHR, and NDCG.

The identification and early treatment of skin diseases are crucial to mitigate serious health risks. The growing attention on researching skin disease analysis stems from the transformative impact of artificial intelligence (AI) in dermatology. In this systematic review, we adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to comprehensively assess recent approaches for skin disease detection. Our study addressed four key research questions exploring the methods for skin disease detection, the evaluation techniques employed to measure the effectiveness of skin disease detection models, the datasets utilized, and the challenges encountered in applying machine learning and deep learning techniques for skin disease detection. We screened studies from 2019 to 2023 from reputable databases, including IEEE Explore, Science Direct, and Google Scholar. Our findings revealed that the CNN model outperformed other deep learning models. Additionally, our analysis identified the ISIC public dataset as the most frequently used dataset. The studies reviewed employed evaluation metrics such as accuracy, recall, precision, sensitivity, and F1 score to evaluate model performance. We identified several limitations in the studies we reviewed, including the use of limited datasets, challenges in distinguishing between diseases with similar features, and other related limitations. Overall, we provided a comprehensive overview of the current state-of-the-art techniques in skin disease detection and highlighted the future directions.

This review explores the integration of Human-Computer Interaction (HCI) principles in AI to advance Human-Centered Artificial Intelligence (HCAI). It highlights how these fields intersect to create user-friendly AI systems that enhance human capabilities and align with human values. Given the recent interest of HCI in user-centered design and AI in technical innovation, this paper bridges this divide by infusing principles from HCI into AI systems. Relevant peer-reviewed articles, conference papers, and case studies have been selected from leading databases like IEEE Xplore, ACM Digital Library, ScienceDirect, and Google Scholar, encompassing publications from 2017 to 2024. The inclusion criteria for the review focus on interdisciplinary approaches, real-world applications, and challenges of HCAI, while studies that do not have a clear methodology or lack relevance to HCAI were excluded. This paper identifies some of the key gaps, highlights the successful applications of HCAI across healthcare, edu-cation, and entertainment, and discusses various challenges that have arisen, such as bias, transparency, and balancing automation with human control. Findings reveal that iterative design and hu-man-centered frameworks will lead to better usability and ethical fit for HCAI, but significant challenges remain. This study proposes an integrative framework for bringing HCI principles into AI design through interdisciplinary collaboration in developing systems that will enhance human capabilities while considering ethical aspects. Future directions include responsible AI, personalized healthcare, and effective human-AI collaboration.

Autonomous mobile robot navigation integrates localization, mapping, and path planning to enable effective operation in complex environments. This study compares a modified Q-learning algorithm with a Deep Q-Network (DQN) in a simulated gym environment, focusing on convergence speed, success rate, and computational efficiency. The modified Q-learning algorithm converged after 44 episodes, outperforming the DQN, which required 400 episodes. It achieved a success rate of 69.6% with a cumulative reward that surpassed the DQN in fewer episodes, while completing simulations in just 9 minutes compared to 400 minutes for the DQN. These results demonstrate the modified Q-learning’s efficiency in addressing the exploration-exploitation trade-off and navigating complex environments. This study highlights the potential of the modified Q-learning algorithm for real-world applications in robotics and autonomous navigation, providing a foundation for future research in intelligent path planning

This paper presents a hybrid quantum-classical machine learning model for classification tasks, integrating a 4-qubit quantum circuit with a classical neural network. The quantum circuit is designed to encode the features of the Iris dataset using angle embedding and entangling gates, thereby capturing complex feature relationships that are difficult for classical models alone. The model, which we term a Quantum Convolutional Neural Network (QCNN), was trained over 20 epochs, achievinga perfect 100% accuracy on the Iris dataset test set on 16 epoch. Our results demonstrate the potential of quantum-enhanced models in supervised learning tasks, particularly in efficiently encoding and processing data using quantum resources. We detail the quantum circuit design, parameterizedgate selection, and the integration of the quantum layer with classical neural network components.This work contributes to the growing body of research on hybrid quantum-classical models and theirapplicability to real-world datasets.

High blood pressure (or hypertension) is a causative disorder to a plethora of other ailments – as it succinctly masks other ailments, making them difficult to diagnose and manage with a targeted treatment plan effectively. While some patients living with elevated high blood pressure can effectively manage their condition via adjusted lifestyle and monitoring with follow-up treatments, Others in self-denial leads to unreported instances, mishandled cases, and in now rampant cases – result in death. Even with the usage of machine learning schemes in medicine, two (2) significant issues abound, namely: (a) utilization of dataset in the construction of the model, which often yields non-perfect scores, and (b) the exploration of complex deep learning models have yielded improved accuracy, which often requires large dataset. To curb these issues, our study explores the tree-based stacking ensemble with Decision tree, Adaptive Boosting, and Random Forest (base learners) while we explore the XGBoost as a meta-learner. With the Kaggle dataset as retrieved, our stacking ensemble yields a prediction accuracy of 1.00 and an F1-score of 1.00 that effectively correctly classified all instances of the test dataset.

Landslides pose significant threats to life, property, and infrastructure. This study explores applying unsupervised learning techniques to identify and understand landslide-prone areas. We analyzed topographic data by employing K-Means, Hierarchical Clustering, Spectral Clustering, Mean Shift Clustering, and DBSCAN to uncover hidden patterns in landslide occurrence. Evaluation metrics, including the Silhouette Score, Davies-Bouldin Index, and Calinski-Harabasz Index, were used to assess the performance of these algorithms. Hierarchical Clustering achieved the highest Silhouette Score of 0.635, indicating excellent cluster separation. However, Mean Shift Clustering outperformed the other methods with a superior Davies-Bouldin Index of 0.603 and the highest Calinski-Harabasz Index of 4121.75, demonstrating the best overall clustering performance. DBSCAN also performed well, with a Silhouette Score of 0.610 and 12 noise points identified. These findings contribute to a deeper understanding of landslide spatial distribution and can inform the development of effective early warning systems and mitigation strategies.

The efficacy of Cancer treatment often varies across different types of cancers. This study aims to investigate any pattern relationship in histopathological images of different cancer types to find any potential correlation between those patterns. Using deep image analysis techniques and artificial intelligence (AI), we extract, analyze, and compare the morphological parameters of cancer images to identify potential indicators of that treatment effective for one type that might be applicable for its correspondence. This research applied advanced image analysis, artificial intelligence (AI), machine learning, and more sophisticated statistical analysis to find the required pattern relationship for those parameters. The study answers the question regarding the correlation of different measurement parameters across different varieties of cancer cells. The model achieved an impressive ROC-AUC score of 0.967, an F1-score of 0.805, and Cohen's kappa coefficient of 0.767, indicating a high level of agreement and predictive performance. The overall accuracy of the model was 81%, with both macro and weighted averages also at 81%. These results provide strong evidence of meaningful pattern relationships across different cancer types, potentially enhancing treatments' applicability across various cancers.

This paper critically examines artificial intelligence in the healthcare sector and aims to identify concrete points of challenges and business value propositions first in radiology and then across healthcare more broadly. It discusses current applications in radiology and future uses of AI in healthcare, focusing on three main areas: (i) emergency incidents handling, (ii) intensive care unit treatment and (iii) augmented telemedicine, to which emergency radiology is a critical success factor. Despite some risks and compliance issues that need to be taken care of, this paper clearly shows that AI has the potential (a) to reengineer the business processes of the healthcare sector, using AI-assisted radiology as a driver and (b) to improve the effectiveness of the healthcare system as well as (c) to increase the quality provision of healthcare services. Despite its slow adoption, AI-assisted healthcare can indeed offer business/operational solutions that benefit all healthcare stakeholders.

The rapid growth in vehicle ownership has led to increased traffic congestion, making the need for autonomous driving solutions more urgent. Autonomous Vehicles (AVs) offer a promising solution to improve road safety and reduce traffic accidents by adapting to various driving conditions without human intervention. This research focuses on implementing Deep Q-Network (DQN) to enhance AV performance in different driving modes: safe, normal, and aggressive. DQN was selected for its ability to handle complex, dynamic environments through experience replay, asynchronous training, and epsilon-greedy exploration. We designed a simulation environment using the Highway-env platform and evaluated the DQN model under varying traffic densities. The performance of the AV was assessed based on two key metrics: success rate and total reward. Our findings show that the DQN model achieved a success rate of 90.75%, 94.625%, and 95.875% in safe, normal, and aggressive modes, respectively. Although the success rate increased with traffic intensity, the total reward remained lower in aggressive driving scenarios, indicating room for optimization in decision-making processes under highly dynamic conditions. This study demonstrates that DQN can adapt effectively to different driving needs, but further optimization is needed to enhance performance in more challenging environments. Future work will focus on improving the DQN algorithm to maximize both success rate and reward in high-traffic scenarios and testing the model in more diverse and complex environments.

Incorporating deep learning models has marked a significant advancement in integrating trends and technology within the fashion industry. These models are extensively applied in the realm of image recognition, product recommendation, and trend prediction, employing deep learning techniques such as Convolutional Neural Networks (CNNs), Generative Adversarial Networks (GANs), Recurrent Neural Networks (RNNs), and Autoencoders. This paper aims to cover various aspects of the textile industry’s supply chain processes, highlighting these deep learning techniques' present influence and potential future directions. It includes a comprehensive analysis of some of the most recent and well-recognized studies in the industry that focus on different parts of a product’s lifecycle in the industry, such as Design and Trend Forecasting, Production and Quality Control, Marketing and Sales, and Distribution and Retail. While deep learning has significantly improved the efficiency of processes across the supply chain, our review highlights some of the existing challenges, such as dependency on large datasets, manual annotation needs, and limitations in creative design generation, encouraging future research to focus on more sophisticated models incorporating multimodal data and personalized factors like body types and aesthetic preferences. Additionally, areas like sewing pattern generation, body-aware designs, and ethical sourcing are critical areas of the fashion industry that require further exploration.

This paper offers a brief analysis highlighting the effectiveness of several well-known Artificial Intelligence (AI) technologies applied in education, particularly in teaching and learning. It provides an overview of how modern classrooms can benefit from more effective teaching strategies that encourage students to engage in hands-on learning. Advanced technologies are changing how knowledge is found and shared, as well as how teaching is delivered. Memorization has been emphasized in educational models as a crucial learning skill until relatively recently. The technologies alter how knowledge is accessed and taught in schools today. Based on that, most knowledge is readily available, quickly accessible, and available online. The skills of reading, sharing, listening, and acting are now prerequisites for schooling. Most recent developments in advanced technologies in education are provided. Some analyses related to opening issues and challenges are shown for future work.

This is one of the greatest challenges in computer network security and cannot be dealt with without a set of most recent detection techniques. This paper lays down a new hybrid technique that combines Clustering-Based Local Outlier Factor (CBLOF) and Extreme Gradient Boosting (XGBoost) to enhance accuracy while detecting Distributed Denial of Service (DDoS) from network traffic. The proposed hybrid model utilizes a CBLOF for outlier detection as feature engineering. Over the detected anomalies, classification is to be done using XGBoost classification to attain the objective. The proposed hybrid model was tested extensively on CICIDS 2017 and CICIDS 2018 datasets Compared with traditional ones, the proposed model outperformed the traditional ones with an accuracy rate of 99.99%, precision of 100%, and F1 score reflecting perfection. These results confirm this model's efficiency in terms of known and novel attack patterns and introduce a further reliable framework for the timely detection of DDoS attacks. Even if it is computation-heavy, optimization could be made towards real-time large-scale data.

Face recognition occluded by occlusions, such as glasses or shadows, remains a challenge in many security and surveillance applications. This study aims to analyze the performance of various machine learning and deep learning techniques in face recognition scenarios with occlusions. We evaluate KNN (standard and FisherFace), CNN, DenseNet, Inception, and FaceNet methods combined with a pre-trained DeepFace model using three public datasets: YALE, Essex Grimace, and Georgia Tech. The results show that KNN maintains the highest accuracy, reaching 100% on two datasets (Essex Grimace and YALE), even in the presence of occlusions. Meanwhile, CNN shows strong performance, with accuracy remaining 100% on YALE, both with and without occlusions, although its performance drops slightly on Essex Grimace (94% with occlusion). DenseNet and Inception show a more significant drop in accuracy when faced with occlusion, with DenseNet dropping from 81% to 72% on Essex Grimace and Inception dropping from 100% to 92% on the same dataset. FaceNet + DeepFace excels on more large dataset (Georgia Tech) with 98% accuracy, but its performance drops dramatically to 53% and 70% on Essex Grimace and YALE with occlusion. These findings indicate that while deep learning methods show high accuracy under ideal conditions, machine learning methods such as KNN are more flexible and robust to occlusion in face recognition.

Each year, more than 1.9 million cases of colorectal cancer (CRC) are diagnosed worldwide. By 2040, the burden of colorectal cancer is estimated to reach 3.2 million new cases per year and 1.6 million deaths per year worldwide. As of 2024, it ranks as the third most prevalent form of cancer, contributing to over 10% of all new cancer cases annually, with a 5-year survival rate of only 65%. With effective early detection mechanisms in place, the survival rate could potentially increase to 90%. However, current detection mechanisms are manual and error-prone.
This study presents a deep learning-based approach to automating the detection of polyps, the tumor that causes colorectal cancer, in the human colon. Various state-of-the-art deep learning models  – including VGG, ResNet, DenseNet, and EfficientNet were trained and tested on a publicly available dataset. The findings of this study show that deep learning models can significantly automate the early diagnosis process of colorectal cancer with high accuracy, especially the DenseNet and EfficientNet models – attaining 99% and 99.4% respectively for both accuracy and F1 score metrics on the test dataset. This study validates the potential of deep learning to enhance the accuracy and reliability of colorectal cancer detection and prevention, ultimately contributing to better quality of diagnosis and patient outcomes.

This research aims to enhance mental well-being by addressing symptoms of anxiety and depression through a personalized, culturally specific multimodal emotion prediction system. It employs an emotionally aware Reinforcement Learning (RL) agent to suggest tailored Cognitive Behavioral Therapy (CBT) activities. The study focuses on developing precise, individualized emotion prediction models using facial expressions, vocal tones, and text, and integrates these models with the RL agent for emotionally aware CBT recommendations. The mHealth approach combines deep learning models with RL, achieving accuracies of 72% for facial expressions, 73% for vocal tones, and 86% for text, all fine-tuned for the Sri Lankan context. Validation through real-world use and user feedback consistently demonstrated that each model exceeds 70% accuracy, fulfilling the objective of precise emotion prediction. A weighted algorithm was introduced to refine the emotion prediction experience and personalize forecasts across the three modalities to enhance mental well-being. The RL-enabled agent suggests CBT activities approved by mental health professionals, tailored based on predicted emotions, and delivered through the same mHealth application. The effectiveness of these interventions was assessed using the DASS-21 questionnaire, revealing significant reductions in depression scores (from 21.08 to 13.54) and anxiety scores (from 19.85 to 10.46) in the study group compared to the control group. The study concludes that integrating multimodal emotion prediction models with RL-based CBT suggestions positively impacts mental well-being and contributes to personalized mental health interventions.

The digital revolution frontiers have rippled across society today – with various web content shared online for users as they seek to promote monetization and asset exchange, with clients constantly seeking improved alternatives at lowered costs to meet their value demands. From item upgrades to their replacement, businesses are poised with retention strategies to help curb the challenge of customer attrition. The birth of smartphones has proliferated feats such as mobility, ease of accessibility, and portability – which, in turn, have continued to ease their rise in adoption, exposing user device vulnerability as they are quite susceptible to phishing. With users classified as more susceptible than others due to online presence and personality traits, studies have sought to reveal lures/cues as exploited by adversaries to enhance phishing success and classify web content as genuine and malicious. Our study explores the tree-based Random Forest to effectively identify phishing cues via sentiment analysis on phishing website datasets as scrapped from user accounts on social network sites. The dataset is scrapped via Python Google Scrapper and divided into train/test subsets to effectively classify contents as genuine or malicious with data balancing and feature selection techniques. With Random Forest as the machine learning of choice, the result shows the ensemble yields a prediction accuracy of 97 percent with an F1-score of 98.19% that effectively correctly classified 2089 instances with 85 incorrectly classified instances for the test-dataset.

Rice is a staple food for most Nigerians, making accurate yield prediction is crucial for food security. This study addresses the limitations of traditional forecasting methods by employing Multiple Linear Regression (MLR) coupled with Recursive Feature Elimination (RFE) to predict rice yield in Adamawa and Cross River states, characterized by distinct agroclimatic conditions. Utilizing climatic data and historical yield records from 1990 to 2022, we trained and evaluated MLR and compared the MLR results with two other machine learning models (XGBoost, and K Nearest Neighbours). RFE-optimized feature selection identified All-sky Photosynthetically Active Radiation (PAR) as a key factor. MLR demonstrated a very stable prediction performance with R² values of 0.90 and 0.92 for Adamawa and Cross River, respectively, after RFE. This research contributes to developing advanced Agro-information systems, supporting informed agricultural decision-making, and enhancing Nigeria's food security.

Malaria continues to pose a significant global health threat, and the emergence of drug-resistant malaria exacerbates the challenge, underscoring the urgent need for new antimalarial drugs. While several machine learning algorithms have been applied to quantitative structure-activity relationship (QSAR) modeling for antimalarial compounds, there remains a need for more interpretable models that can provide insights into the underlying mechanisms of drug action, facilitating the rational design of new compounds. This study develops a QSAR model using Light Gradient Boosting Machine (LightGBM). The model is integrated with SHapley Additive exPlanations (SHAP) to enhance interpretability. The LightGBM model demonstrated superior performance in predicting antimalarial activity, with an ac-curacy of 86%, precision of 85%, sensitivity of 81%, specificity of 89%, and an F1-score of 83%. SHAP analysis identified key molecular descriptors such as maxdO and GATS2m as significant contributors to antimalarial activity. The integration of LightGBM with SHAP not only enhances the predictive ac-curacy of the QSAR model but also provides valuable insights into the importance of features, aiding in the rational design of new antimalarial drugs. This approach bridges the gap between model accuracy and interpretability, offering a robust framework for efficient and effective drug discovery against drug-resistant malaria strains.

Phishing attacks continue to be a significant threat to internet users, necessitating the development of advanced detection systems. This study explores the efficacy of a Bidirectional Gated Recurrent Unit (BiGRU) model combined with feature selection techniques for detecting phishing websites. The dataset used for this research is sourced from the UCI Machine Learning Repository, specifically the Phishing Websites dataset. This approach involves cleaning and preprocessing the data, then normalizing features and employing feature selection to identify the most relevant attributes for classification. The BiGRU model, known for its ability to capture temporal dependencies in data, is then applied. To ensure robust evaluation, we utilized cross-validation, dividing the data into five folds. The experimental results are highly promising, demonstrating a Mean Accuracy, Mean Precision, Mean Recall, Mean F1 Score, and Mean AUC of 1.0. These results indicate the model's exceptional performance distinguishing between phishing and legitimate websites. This study highlights the potential of combining BiGRU models with feature selection and cross-validation to create highly accurate phishing detection systems, providing a reliable solution to enhance cybersecurity measures.

Segmentation of pneumonia areas on chest X-rays is essential to improve the accuracy of recognition tasks and subsequent diagnosis. The capabilities of deep learning techniques, U-Net, SegNet, and DeepLabV3, are assessed to achieve these purposes. Using transfer learning, these models were adapted to pneumonia-specific datasets. The evaluation focuses on Intersection over Union (IoU) and accuracy metrics. Results show that DeepLabV3 outperforms U-Net and SegNet, achieving 84.4% accuracy and 81% IoU. U-Net achieves 80.3% accuracy and 68% IoU, while SegNet achieves 81.0% accuracy and 70% IoU. These findings highlight the potential of transfer learning models to automate the segmentation of pneumonia-affected regions, thereby facilitating timely and accurate medical intervention.

This research aims to evaluate and compare the performance of several deep learning architectures, especially InceptionV3 and InceptionResNetV2, with other models, such as EfficientNetB3, ResNet50, and VGG19, in classifying rice leaf diseases. In addition, this research also evaluates the impact of using data augmentation on model performance. Three different datasets were used in this experiment, varying the number of images and class distribution. The results show that InceptionV3 and InceptionResNetV2 consistently perform excellently and accurately on most datasets. Data augmentation has varying effects, providing slight advantages on datasets with lower variation. The findings from this research are that the InceptionV3 model is the best model for classifying rice diseases based on leaf images. The InceptionV3 model produces accuracies of 99.53, 58.94, and 90.00 for datasets 1, 2, and 3, respectively. It is also necessary to be wise in carrying out data augmentation by considering the dataset's characteristics to ensure the resulting model can generalize well.

This research explores the effectiveness of machine learning (ML) and deep learning (DL) classification methods in Aspect-Based Sentiment Analysis (ABSA) on product reviews, incorporating Latent Dirichlet Allocation (LDA) for topic modeling. Using the Amazon reviews dataset, this research tests models such as Naive Bayes (NB), Support Vector Machines (SVM), Random Forest (RF), Long Short-Term Memory (LSTM), and Gated Recurrent Units(GRU). Important aspects such as the product's quality, practicality, and reliability are discussed. The results show that the RF and DL models provide competitive performance, with the RF achieving an accuracy of up to 94.50% and an F1 score of 95.45% for the reliability aspect. The study's conclusions emphasize the importance of selecting an appropriate model based on specifications and data requirements for ABSA, as well as recognizing the need to strike a balance between accuracy and computational efficiency.

In the clinical treatment of skin diseases and cancer, cryotherapy and immunotherapy offer effective and minimally invasive alternatives. However, the complexity of patient response demands more sophisticated analytical strategies for accurate outcome prediction. This research focuses on analyzing the effect of preprocessing in various machine learning models on the prediction performance of cryotherapy and immunotherapy. The preprocessing techniques analyzed are advanced feature engineering and Synthetic Minority Over-sampling Technique (SMOTE) and Tomek links as resampling techniques and their combination. Various classifiers, including support vector machine (SVM), Naive Bayes (NB), Decision Tree (DT), Random Forest (RF), XGBoost, and Bidirectional Gated Recurrent Unit (BiGRU), were tested. The findings of this study show that preprocessing methods can significantly improve model performance, especially in the XGBoost model. Random Forest also gets the same results as XGBoost, but it can also work better without significant preprocessing. The best results were 0.8889, 0.8889, 0.6000, 0.9037, and 0.8790, respectively, for accuracy, recall, specificity, precision, and f1 on the Immunotherapy dataset, while on the Cryotherapy dataset, respectively, they were 0.8889, 0.8889, 0.6000, 0.9037, and 0.8790. This study confirms the potential of customized preprocessing and machine learning models to provide deep insights into treatment dynamics, ultimately improving the quality of diagnosis.

This research aims to develop a robust diabetes classification method by integrating the Synthetic Minority Over-sampling Technique (SMOTE)-Tomek technique for data balancing and using a machine learning ensemble led by eXtreme Gradient Boosting (XGB) as a meta-learner. We propose an ensemble model that combines deep learning techniques such as Bidirectional Long Short-Term Memory (BiLSTM) and Bidirectional Gated Recurrent Units (BiGRU) with XGB classifier as the base learner. The data used included the Pima Indians Diabetes and Iraqi Society Diabetes datasets, which were processed by missing value handling, duplication, normalization, and the application of SMOTE-Tomek to resolve data imbalances. XGB, as a meta-learner, successfully improves the model's predictive ability by reducing bias and variance, resulting in more accurate and robust classification. The proposed ensemble model achieves perfect accuracy, precision, recall, specificity, and F1 score of 100% on all tested datasets. This method shows that combining ensemble learning techniques with a rigorous preprocessing approach can significantly improve diabetes classification performance.

This research aims to improve the accuracy of liver disease classification using Quantum Feature Engineering (QFE) and the Synthetic Minority Over-sampling Tech-nique and Tomek Links (SMOTE-Tomek) data balancing technique. Four machine learning models were compared in this research, namely eXtreme Gradient Boosting (XGB), Random Forest (RF), Support Vector Machine (SVM), and Logistic Regression (LR) on the Indian Liver Patient Dataset (ILPD) dataset. QFE is applied to capture correlations and complex patterns in the data, while SMOTE-Tomek is used to address data imbalances. The results showed that QFE significantly improved LR performance in terms of recall and specificity up to 99%, which is very important in medical diagnosis. The combination of QFE and SMOTE-Tomek gives the best results for the XGB method with an accuracy of 81%, recall of 90%, and f1-score of 83%. This study concludes that the use of QFE and data balancing techniques can improve liver disease classification performance in general.