Applying Machine Learning to Predict Software Defects

The quality of software is its most important attribute. Over the past few years, software defect prediction has been increasingly popular and could have a direct impact on quality. Software quality is greatly reduced by defective software modules, which leads to budget overruns, missed deadlines, and far higher maintenance costs. Software complexity also leads to an increase in flaws, making human defect detection a laborious procedure. This provided motivation for academics to create methods for automatically identifying software flaws. The goal of this investigation was to improve the dataset's precision and accuracy of the predictive algorithm in relation to past research. To achieve this, we classified the class labels using K-means clustering. We also used classification models on a subset of the features. ML models are optimised by the use of particle swarm optimisation. The classifiers' efficiency was evaluated using a confusion matrix, f-measure, precision, accuracy, recall, and performance error measures. All ML and optimised ML models, according to the results, produce the best outcomes; however, the ANN and optimised ANN models perform best, achieving the maximum accuracy of 99.3% and 99.9%, respectively.