OpEffiRes Net: Optimization based EfficientNetB0-ResNet50 and correlation based feature selection for software...

Software Failure Prediction (SFP) includes estimating the probability that a software system will fail during its operational period. Its primary goal is to recognize components that are prone to defects or time intervals when failures are more probable, allowing the implementation of preventive measures and thus augmenting overall reliability of software. The study aims to create an advanced hybrid Deep Learning (DL) strategy for SFP. Experiments were conducted using the Software Defect Prediction dataset obtained from Kaggle. The procedure begins with raw input data undergoing a pre-processing stage, where it is cleaned and transformed into a format suitable for accurate feature extraction. Min-max normalization is applied to scale the data consistently. Next, the most relevant features are selected by employing proposed Double Exponential-Coati Optimization Algorithm (DE-COA), a combination of Coati Optimization Algorithm (COA) and Double Exponential Smoothing with correlation analysis to enhance feature selection. For the prediction, a novel approach, called OpEffiResNet model, is presented by combining EfficientNetB0 and ResNet50. EfficientNetB0 is selected for its ability to achieve high accuracy with a negligible parameter count and reduced computational cost, while ResNet50 offers a deep residual architecture known for its robustness and stability. The integration of these models enables the proposed model to deliver accurate, efficient, and reliable software fault predictions. Model performance was evaluated using 5-fold cross-validation, and results are reported as mean ± standard deviation to ensure statistical reliability. The proposed approach achieved an average accuracy of 98.8% ± 0.6, sensitivity of 99.4% ± 0.4, specificity of 99.0% ± 0.5, precision of 98.8% ± 0.7, and F1-score of 98.7% ± 0.6. The results demonstrate that the proposed framework can effectively improve software defect prediction performance. Future work will focus on validating the model on multiple datasets and comparing it with advanced tabular learning approaches.