A Systematic Approach for Enhancing Software Defect Prediction Using Machine Learning

In the modern world of software development, ensuring reliability and performance is of paramount importance. However, despite the best efforts from the developers, software defects can still emerge, causing frustration and wasted resources. Due to the numerous defects found during the software development process, researchers have developed numerous ways for defect prediction models. However, these models cut down the time and expense of development when problems in a concurrent software product are anticipated. Due to the increased amount of defects brought on by software complexity, manual defect detection can become an extremely time-consuming procedure. This encouraged researchers to create methods for the automatic detection of software defects. The study of this paper has shown that a combination of machine learning algorithms could be applied effectively for software defect prediction. Interestingly, the combination of Artificial Neural Network and Random Forest classifier has been performed with the mean accuracy of 91%, while the hyper-parameter optimization model classifier has been performed with the mean accuracy of 83%, 83%, 84%, 77% and 80% for Support Vector Machine, Random Forest, Logistic Regression, Naive Bayes Gaussian and Decision Tree, respectively. These findings have demonstrated the potential of Machine Learning in the area of software development.