Software Defect Prediction Model Based on Black-Faced Arachnid Optimization Using Deep Long Short-Term Memory Classifier
M. Prashanthi, M Chandramohan · Web Intelligence · 2026
Software Defect Prediction is a critical challenge in programming language research and software development aimed at enhancing software quality and reliability. Accurately identifying faulty source code can be difficult, given the complexity of the task. Existing software defect prediction models, designed to estimate the precise number of faults in software applications, failed to deliver precise results due to the presence of noisy input data. Thus, the research proposes the Black-Faced Arachnid Optimized Deep Long Short-Term Memory (LSTM) model for the accurate detection of software defects. In essence, the model leverages the deep LSTM to capture relevant patterns over long sequences and to mitigate the vanishing gradient problems effectively. Additionally, the research proposes the Black-Faced Arachnid (BFA) optimization algorithm to perform an effective hyperparameter tuning process. Moreover, the BFA-optimized feature selection process is introduced to effectively identify the highly relevant features related to defects and non-defects. Furthermore, the experimental results demonstrate that the model achieved a greater accuracy of 98.54%, F1-score of 98.64%, precision of 98.97%, sensitivity of 98.32%, and specificity of 98.81%, for a training percentage of 90 using the PROMISE Dataset, respectively.