Software testing is a critical phase of the software development lifecycle, yet identifying defect-prone modules remains challenging due to increasing system complexity. Traditional test automation focuses on executing test cases but lacks intelligence in prioritizing high-risk components. This paper investigates the application of machine learning techniques for software defect prediction using static code metrics. A Logistic Regression model was trained and evaluated on a publicly available software defect dataset. Experimental results demonstrate that addressing class imbalance significantly improves the detection of defective modules, particularly recall, which is crucial for effective testing. The findings highlight the potential of integrating AI-driven defect prediction into test automation to support risk-based testing and improve overall testing efficiency.
Dhruv Sharma (Wed,) studied this question.