A Data Driven Approach to Software Quality Estimation Using Modern Machine Learning Techniques

Striving for software quality presents elevated obstacles in modern development environments which contain large sophisticated codes that undergo frequent updates. Traditional quality assurance methods—such as manual reports and static analysis—are often time-taking process, subjective, and not scalable thing. This research develops a data-based system for quality assessment of software using contemporary machine learning methods. Multiple software metrics consisting of code complexity along with churn rate and defect density and processrelated factors are gathered by the system to conduct analysis. The system processes real repository data from GitHub together with approved defect datasets to teach supervised ML algorithms Random Forest, XGBoost, and Support Vector Machines (SVM) for predicting faulty modules. The proposed automation solution delivers real-time evaluation of software quality using usercentric dashboards that provide concrete insights to developers. Model performance evaluation depends on accuracy measures and precision among other metrics that include recall and AUCROC. The system is designed to work within CI/CD pipelines and it improves both the reliability and maintainability of software development projects. Modern software quality assurance stands to gain an efficient scalable intelligent solution through this approach.