Microfinance institutions (MFIs) are pivotal to financial inclusion in emerging economies, yet they face heightened credit risk due to borrower informality, data scarcity, and severe class imbalance. Motivated by the microfinance context, this study proposes a human-centered hybrid machine learning framework that integrates ensemble learning with Synthetic Minority Over-sampling Technique (SMOTE) to enhance default detection while supporting transparent and responsible decision-making. Using a large-scale public credit application dataset as an empirical benchmark, we compare logistic regression, Random Forest, AdaBoost and Naïve Bayes models under imbalanced and rebalanced conditions. The results indicate that class rebalancing substantially improves minority-class detection, with the Random Forest + SMOTE configuration achieving the best performance (F1 = 0.73; AUC = 0.97). Beyond predictive accuracy, the findings highlight the importance of human oversight and explainability to mitigate exclusionary risks. The study offers practical guidance for MFIs seeking to leverage artificial intelligence while preserving financial sustainability and social inclusion objectives.

This study investigates whether annual report readability influences the project cost of equity capital (COE) in the emerging market of Egypt. To reassess this relationship within a project evaluation context, the research develops a novel, explainable heterogeneous ensemble model capable of capturing complex nonlinear interactions among financial and textual determinants affecting COE estimation. The proposed ensemble integrates Gradient Boosting Regression (GBR), Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Elastic Net using an average-voting strategy. Across multiple evaluation metrics, the ensemble model outperforms linear regression benchmarks, homogeneous bagging and boosting ensembles, and its individual base learners. Specifically, the model achieves superior predictive performance, with an R² of 0.2538, Mean Squared Error (MSE) of 0.0078, Mean Absolute Error (MAE) of 0.0656, and Root Mean Squared Error (RMSE) of 0.0885, significantly outperforming both linear regression and state-of-the-art machine learning alternatives. Feature importance analysis using RF shows that the market-to-book ratio (MTB) and return on equity (ROE) contribute most to predictive accuracy (18.1% and 18.7%, respectively), highlighting the dominant role of financial fundamentals in project COE estimation. Conversely, readability measures exhibit minimal influence. Shapley Additive Explanations (SHAP) further confirm that annual report readability does not exert a statistically meaningful impact on COE within the Egyptian context. By leveraging advanced machine learning and explainability techniques, this study enhances understanding of COE determinants and offers evidence-based insights to improve project appraisal, financial planning, and strategic decision-making.

A major obstacle to breast cancer detection in ultrasound images is that they can be difficult to evaluate due to speckle noise and low contrast as well as very subtle differences between the appearance of benign vs. malignant lesions. With the advent of large amounts of digital medical imaging, particularly ultrasound, there exists a huge volume of heterogeneous ultrasound data; thus, it is imperative to develop both robust big data analytical frameworks as well as artificial neural network (ANN) architectures that will allow the identification of clinically relevant characteristics from these high dimensional noisy data streams. This paper proposes a new dynamic feature fusion technique called adaptive gated cross attention fusion (AGCAF). AGCAF utilizes a learnable gating function and two-way cross attention to dynamically weight each contribution of features from the DeIT and ViT backbones. The AGCAF framework includes a rigorous multiple stage image processing pipeline that uses CLAHE, anisotropic diffusion based speckle removal, and bilateral edge preserving filtering to make the boundaries of the lesions visible before the ANNs extract features. Tools used to visualize and validate the behavior of the model include T-SNE embedding, GRAD-CAM saliency maps, and roc curves. Extensive experiments using the BUSI, BUS-BRA, BrEaST, and BUSI_WHU databases show that the AGCAF architecture provides accuracy rates of 95.18 % and AUC rates of 96.02 % for 3 class classifications, and accuracy and AUC rates of 97.45 % and 97.43 % respectively for binary classifications. Results also show that AGCAF performed better than static concatenation ensemble and all baseline models individually. Additionally, cross database results and ablation studies were conducted to verify the robustness and clinical validity of the proposed method.

Accurate Remaining Useful Life (RUL) prediction is essential for reliable battery management and cost-effective maintenance in lithium-ion energy storage applications. This study proposes a leakage-safe learning framework for battery RUL prediction from numerical cycling features, emphasizing realistic generalization to unseen batteries through group-aware splitting. The core contribution is a proper out-of-fold (OOF) stacking ensemble trained under GroupKFold to prevent information leakage across samples from the same battery while enabling an effective meta-learner to combine diverse learners. Experiments are conducted on a public Battery RUL dataset from Kaggle, and performance is evaluated using complementary regression metrics (R², MAE, RMSE) and robust percentage errors (sMAPE, WAPE). Results show that the proposed stacking approach (OOF GroupKFold with Meta-XGBoost) achieves the best overall performance (R² = 0.999550, MAE = 5.297907, RMSE = 6.834146, sMAPE = 3.080569, WAPE = 0.958618), outperforming strong baselines including XGBoost and Random Forest. These findings confirm that leakage-safe group-aware stacking can significantly enhance accuracy and stability for battery RUL prediction in practical deployment settings.

In recent years, the adoption of advanced machine learning techniques has revolutionized approaches to solving complex problems, such as identifying occurrences of forest fires. Among these techniques, the use of Convolutional Neural Networks (CNNs) combined with ensemble methods is particularly promising. To investigate the feasibility of detecting fires using video streams from Unmanned Aerial Vehicles (UAVs), the lightweight CNN architecture MobileNetV2 was utilized for real-time detection. Several experiments were conducted on the DeepFire dataset, which comprises an equal number of images with and without fire, to evaluate MobileNetV2's performance. Notably, the architecture's linear bottlenecks and the efficient use of inverted residuals ensure high accuracy without compromising on feature extraction capabilities. For a comprehensive assessment, MobileNetV2 was benchmarked against other models, including DenseNet121, EfficientNetV2S, and VGG16. Accuracy was enhanced by averaging predictions through methods such as voting or summing results. As documented in the literature, MobileNetV2 consistently outperforms other architectures in computational efficiency and provides an excellent balance between efficiency and the quality of learned features over multiple epochs. This study underscores the suitability of MobileNetV2 for real-time applications on drones, particularly for the detection of forest fires in resource-constrained environments. The results show that MobileNetV2 achieves the highest accuracy (0.994), sensitivity (0.994), and specificity (0.998) among the tested models, with low standard deviations across all metrics. In contrast, EfficientNetV2S exhibited the lowest accuracy and sensitivity, both at 0.779, with a specificity of 0.829. The ensemble (Sum) method achieved an average accuracy of 0.989, sensitivity of 0.989, and specificity of approximately 0.988. Therefore, MobileNetV2 not only delivers the highest accuracy and stability but also demonstrates that the choice of ensemble method significantly affects the results.

The increasing amount of IoT devices increases the size of network traffic data, causing an increase in the incidence of security breaches in IoT networks. Cybercriminals have developed malware to compromise the security of sensitive data, among other cyber threats. In the presence of inadequate and robust security mechanisms, sensitive data is prone to vulnerability. Hence, protecting data in the IoT environment is becoming a mandatory task. Various approaches have addressed malware detection using network data features. However, there is still room for improvement in developing superior techniques and utilizing more comprehensive datasets. This paper presents a novel lightweight ensemble voting classifier to detect malware traffic by deploying the best possible network data. The merits of the correlation coefficient and Opposition-Based Crow Search Algorithm (OCS) have been leveraged to compute the best possible features. Another advantage of this proposed experiment is its focus on a dataset tailored to malware traffic features. This focus enables highly accurate malware detection. After feature selection using OCS, the proposed malware classifier is trained and validated with both 5-fold and 10-fold cross-validation techniques. The tested results confirm that the presented malware classifier performs best using a minimal feature set, which is highly advantageous for IoT networks due to resource constraints.

Dental caries is arguably the most persistent dental condition that affects most people over their lives. Carious lesions are commonly diagnosed by dentists using clinical and visual examination along with oral radiographs. In many circumstances, dental caries is challenging to detect with photography and might be mistaken as shadows for various reasons, including poor photo quality. However, with the introduction of Artificial Intelligence and robotic systems in dentistry, photographs can be a helpful tool in oral epidemiological research for the assessment of dental caries prevalence among the population. It can be used particularly to create a new automated approach to calculate DMF (Decay, Missing, Filled) index score. In this paper, an autonomous diagnostic approach for detecting dental cavities in photos is developed using deep learning algorithms and ensemble methods. The proposed technique employs a set of pretrained models including Xception, VGG16, VGG19, and DenseNet121 to extract essential characteristics from photographs and to classify images as either normal or caries. Then, two ensemble learning methods, E- majority and E-sum, are employed based on majority voting and sum rule to boost the performances of the individual pretrained model. Experiments are conducted on 50 images with data augmentation for normal and caries images, the employed E-majority and E-sum achieved an accuracy score of 96% and 97%, respectively. The obtained results demonstrate the superiority of the proposed ensemble framework in the detection of caries. Furthermore, this framework is a step toward constructing a fully automated, efficient decision support system to be used in the dentistry area.

Phishing websites are characterized by distinguished visual, address, domain, and embedded features, which identify and defend such threats. Yet, phishing website detection is challenged by overlapping these features with legitimate websites’ features. As the inter-class variance between legitimate and phishing websites becomes low, commonly utilized machine learning algorithms suffer from low performance in overlapping feature cases. Alternatively, ensemble learning that combines multiple predictions intending to address low inter-class variations in the classified data improves the performance in such cases. Ensemble learning utilizes multiple classifiers of similar or different types with multiple deviations of the training data. This paper develops a framework based on random forest ensemble techniques. The limitations of the random forest are the inability to capture the high correlation between features and their join dependency on the label. The random forest is combined with k-means clustering to capture the feature correlation. The framework is evaluated for phishing detection with a dataset of 5000 samples. The results showed the proposed framework over-performed the random forest classifier, all other ensemble classifiers, and the conventional classification algorithms. The proposed framework achieved an accuracy of 98.64%, precision of 0.986, recall of 0.987, and F-measure of 0.986.

The rapid growth of healthcare information systems has led to an increase in the volume of clinical data stored in electronic medical record systems. This data contains valuable information such as patient demographics, vital signs, diagnoses, and medical procedures, which can be used to support clinical decision-making. Determining the appropriate clinical procedure remains a challenge due to the complexity of patient conditions and the potential need for multiple procedures simultaneously. This study aims to develop a robust model for predicting clinical procedures using a multi-label classification approach based on electronic medical record data. The proposed methodology integrates data preprocessing, feature engineering, followed by the implementation of several Machine Learning methods such as LightGBM, XGBoost, CatBoost, Deep Learning Models and combined using an Ensemble Learning method. Various Ensemble Learning techniques, including Simple Average, Machine Learning Average, and Optimized Weight-based combinations, are applied to improve predictive performance. Evaluation is performed using the Micro F1 score and Macro F1 score to assess overall performance. Experimental results show that individual models achieve competitive performance, with LightGBM, XGBoost, CatBoost outperforming the deep learning approach on tabular clinical data. The Ensemble Learning approach further improved performance, with the Optimized Weight with Grid Search Optimization ensemble achieving the highest Micro F1 score of 0.7788 and Macro F1 score of 0.7674. These results also demonstrate that combining multiple models effectively reduces bias and variance while improving generalization. The small difference between Micro and Macro F1 scores indicates balanced performance across labels. In conclusion, the proposed Ensemble Learning-based multi-label prediction model demonstrates robust capabilities in handling complex clinical data and improving prediction accuracy. This study highlights the importance of model combination strategies rather than increasing model complexity. These findings have practical implications for supporting clinical decision-making systems in healthcare settings.

Despite the growing adoption of artificial intelligence (AI) in healthcare, improving AI frameworks remains essential for effective medical implementation. A major challenge in developing reliable machine learning models is handling imbalanced medical datasets. Although metaheuristic algorithms can improve machine learning performance, their effectiveness on imbalanced medical data has received limited attention. This study evaluates four nature-inspired optimization techniques, bat algorithm (BA), genetic algorithm (GA), improved grey wolf optimization (IGWO), and improved whale optimization (IWO), for enhancing balanced random forest (BRF) classification on imbalanced datasets involving breast tumors, heart failure survival, mammogram microcalcification detection, and diabetes. Based on F1 scores, IGWO achieved the greatest improvements (0.256–0.060), followed by IWO (0.229–0.045), GA (0.214–0.036), and BA (0.203–0.036). Validation results confirmed IGWO’s strong convergence and optimization efficiency, consistently outperforming the other methods. The best performance was achieved by IGWO-BRF in breast cancer diagnosis (F1 = 1.000), while the lowest was observed for BA-BRF in mammogram microcalcification detection (F1 = 0.570). Optimization time depended on both the search strategy and dataset characteristics, with IWO showing the fastest optimization across all datasets. Overall, nature-inspired optimizers significantly improved classification performance on imbalanced medical datasets.