Ingegneria del Software 2 Project-1 - Univerity of Tor Vergata

This project presents a comprehensive analysis of defect prediction in software systems using ML classifiers. I have conducted a what-if analysis on two Apache projects (BOOKKEEPER and SYNCOPE) to evaluate the potential impact of refactoring actionable features on bug prevention. My methodology involved dataset generation from Git and JIRA histories, feature extraction at method-level gran- ularity, classifier training and evaluation, and simulation of refactoring scenarios. The results (see: Report_ISW2_DiPalmaFlavio.pdf) demonstrate that reducing actionable features can lead to a reduction in predicted defects, with drop rates of 18.16% for BOOKKEEPER and 21.42% for SYNCOPE. The codebase maintains zero code smells on SonarCloud, ensuring high code quality