Exploring Advanced Techniques and Enhancing Software Quality Assurance Through Machine Learning-based Fault Prediction

In pursuit of impeccable software quality, crucial for ensuring customer satisfaction and economizing testing efforts, a comprehensive examination of diverse machine learning (ML) techniques was undertaken. Leveraging both established and optimized ML methodologies on an openly accessible dataset, our research aimed at enhancing model performance, particularly in terms of accuracy and precision, surpassing preceding studies. Notably, K-means clustering was employed for class label categorization, followed by the application of classification models on discerned features. Particle Swarm Optimization was instrumental in refining ML models. In our evaluation, we looked at various factors such as precision, recall, F-measure, and different performance error metrics, as well as using a confusion matrix. Our findings showed that both regular machine learning models and enhanced versions performed at their best. Particularly, SVM and its enhanced version achieved high accuracy, with the rates of 99.20% and 99.91%, respectively. The corresponding accuracy rates for NB, RF and the ensemble were it is also impressive with percentages of 94.62, 98.82, and 99%, respectively strong performance. Additionally, the enhanced versions of NB and RF achieved accuracy rates of 94.62% and 99.72%, respectively.