Abstract

Software defect is a significant area of software development that has called for the attention of stakeholders in the last few decades. The more software evolves, the more its reliability is of paramount. The quest to predict defects or better still produce error-free software is commendable, however, other factors such as identifying the best LM for prediction and time taken for prediction called for imperative attention. The prolonged processing of dataset in prediction can lead to misclassification. The long processing of dataset is inevitable when the large dataset is used in prediction. That is why this study has applied a meta-heuristic optimization algorithm for feature selection, five classifiers- Support Vector Machine (SVM), Artificial Neural Network (ANN), K-Nearest Neigbour (KNN), Naïve Baiye (NB), and C4.5, four datasets; Mylyn, Eclipse, Lucene, and Equinox and five evaluation metrics- precision, accuracy, recall, classification time, and F1 score. The operational output of the model for prediction was developed and achieved with all the aforementioned tools. The recorded results with HSA revealed that the ANN algorithm achieved the lowest classification time of 24.09s in the Eclipse Dataset which shows that the predictive rate of ANN outperformed other classifiers used for defect classification.