XTREE: An AI-Driven Framework for Planning Effective Software System Improvements

Abstract

Business decision-makers increasingly demand actionable insights from analytics, moving beyond predictive analysis to decision-making support. This paper presents XTREE, an artificial intelligence (AI)-based decision support framework for software projects. XTREE employs supervised machine learning, specifically decision tree algorithms, to learn and recommend plans that improve software metrics such as defect counts and runtime efficiency. By analyzing datasets with weighted class labels indicating quality, XTREE suggests precise changes to input features to transition a software system from a “bad” to a “better” quality state. Evaluated on 11 datasets across various software projects, XTREE outperformed three state-of- the-art planning methods, achieving median improvements of up to 56% and maximum improvements of 77% in defect reduction and runtime optimization, respectively. This research highlights XTREE’s effectiveness in delivering interpretable and actionable plans, making it a valuable tool for AI applications in software engineering and business analytics.