A Machine Learning Approach for Bug or Error Prediction using Cat-Boost Algorithm
Sachin H. Darekar, Prachi Nilekar, Shweta Lilhare, Ashvini Chaudhari, R Narayan, Vishal Borate
Abstract
Quality assurance is an inherent part of the software development life cycle, and bug prediction has a huge role in the reduction of costs and ensures reliable delivery. In this work, a high- performance, gradient-boosting algorithm, CatBoost, applied to software bug prediction. More specifically, the CatBoost model can efficiently deal with categorical features and provide robustness to overfitting using a dataset containing historical software metrics and defect data. It will thus be run against traditional machine learning algorithms and will spot probable bugs with a higher accuracy, precision, and recall. Hence, it is expected to indicate with a good level of accuracy the most important predictors of software defects and thus give insights that can help in integration into development practices geared at early bug detection. That is, CatBoost is one of the most powerful tools for predicting software bugs and thus provides ways toward more efficient and reliable software development processes.