SCALABLE AND ADAPTIVE MACHINE LEARNING MODELS FOR EARLY SOFTWARE FAULT PREDICTION IN AGILE DEVELOPMENT: ENHANCING SOFTWARE RELIABILITY AND SPRINT PLANNING EFFICIENCY

Abstract

This research demonstrates to development and evaluation of scalable, adaptive machine learning models for early fault prediction in Agile software development. This research addresses the critical challenge of late-stage fault detection in Agile software development, which leads to costly rework and reduced code quality. The proposed research involves the new fault forecasting framework that features the combination of both scalability and adaptability, which are scarcely implemented in the existing models, but which is specifically designed to be adapted to a dynamic Agile environment. Our model, in contrast to the conventional static models, is continuously changing utilizing online learning according to real-time sprint data it receives. It is important in an Agile environment whereal code change is commonand allows finding faults much faster and more accurate, which helps plan sprints better, code better structure, and software reliability. Integration into Agile workflows resulted in a 27% boost in programmer productivity and 21% enhancement in sprint planning efficiency. The findings demonstrate the practical value of integrating adaptive ML models to deliver higher-quality software faster and more reliably. This work shows the practical utility of ML-driven fault forecast in offering high-quality software more efficiently in dynamic development settings.