Software defect prediction identifies defect-prone modules before testing, reducing costs and development time. Machine learning techniques are widely used, but high-dimensional datasets often degrade classification accuracy due to irrelevant features. To address this, effective feature selection is essential but remains an NP-hard challenge best tackled with heuristic algorithms. This study introduces a binary, multi-objective starfish optimizer for optimal feature selection, balancing feature reduction and classification performance. A Choquet fuzzy integral-based ensemble classifier further enhances prediction reliability by aggregating multiple classifiers. The approach was validated on five NASA datasets, demonstrating superior performance over traditional classifiers. Key software metrics—such as design complexity, operators and operands count, lines of code, and numbers of branches—were found to significantly influence defect prediction. The results show that the proposed method improves classification performance by 1% to 13% while retaining only 33% to 57% of the original feature set, offering a reliable and interpretable solution for software defect prediction. This approach holds strong potential for broader, high-dimensional classification tasks.

Software defect prediction aims to identify defect-prone modules before testing, reducing costs and duration. Machine learning (ML) techniques are widely used to develop predictive models for classifying defective software components. However, high-dimensional training datasets often degrade classification accuracy and precision due to irrelevant or redundant features. To address this, effective feature selection is crucial, but it poses an NP-hard challenge that can be efficiently tackled using heuristic algorithms. This study introduces a Binary Multi-Objective Starfish Optimizer (BMOSFO) for optimal feature selection, enhancing classification accuracy and precision. The proposed BMOSFO balances two conflicting objectives: minimizing the number of selected features and maximizing classification performance. A Choquet Fuzzy Integral-based Ensemble Classifier is then employed to further enhance prediction reliability by aggregating multiple classifiers. The effectiveness of the proposed approach is validated using five real-world NASA benchmark datasets, demonstrating superior performance compared to traditional classifiers. Experimental results reveal that key software metrics—such as design complexity, operators and operands count, lines of code, and number of branches—significantly influence defect prediction. The findings confirm that BMOSFO not only reduces feature dimensionality but also enhances classification performance, providing a robust and interpretable solution for software defect prediction. This approach shows strong potential for generalization to other high-dimensional classification tasks.

Human Activity Recognition (HAR) plays a crucial role in the development of intelligent systems for smart healthcare, sustainable living environments, and enhanced human well-being. However, conventional machine learning (ML) and deep learning (DL) approaches often face challenges in handling noisy real-world sensor data or require high computational resources. To address these limitations, this study proposes a novel two-stage hybrid ensemble framework that integrates probabilistic outputs from multiple ML classifiers with advanced DL models for robust HAR. In the first stage, diverse ML classifiers, including K-Nearest Neighbors (KNN), Decision Tree (DT), Naive Bayes (NB), Logistic Regression (LR), and Support Vector Machines (SVM) are independently trained to generate class probability vectors. These outputs are then fused to construct a meta-feature representation. In the second stage, deep learning models such as Multi-Layer Perceptron (MLP), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Bi-directional Long Short-Term Memory (BiLSTM), and Autoencoder are trained on the meta-feature space to perform final classification. The proposed framework is evaluated on the benchmark Human Activity Recognition Using Smartphones dataset and achieves superior performance with an average accuracy of 99.21%, precision of 99.26%, recall of 99.24%, and F1-score of 99.25%. Statistical validation confirms the robustness and generalization capability of the model compared to state-of-the-art approaches. By enabling accurate, efficient, and scalable activity recognition, the proposed system contributes to the advancement of intelligent healthcare solutions, remote patient monitoring, and inclusive digital ecosystems. This work supports the development of sustainable and human-centric intelligent systems that enhance quality of life and promote well-being in real-world environments.