Predicting Software Defects Using Machine Learning and the NASA JM1 Dataset

Program flaws can lower system reliability and raise maintenance costs, making them a significant challenge in program development. Early defect detection helps developers concentrate on important modules and enhance program quality. This study uses the NASA JM1 dataset to provide a machine learning-based method for software defect prediction. The dataset contains a variety of software metrics that are utilised as input features for the prediction model, including lines of code, complexity metrics, operator counts, and branch counts. A Random Forest classifier is used in this study to evaluate these metrics and categorise software modules as either non-defective or defective. A web-based application created with the Django framework incorporates the trained model, enabling users to enter software measurements and receive real-time fault predictions. According to experimental data, the suggested method can successfully identify modules that are prone to defects, assisting developers in setting testing priorities and enhancing overall program reliability. Through intelligent defect prediction, this system offers a workable way to support software quality assurance.