A Voting Classifier Framework for Improved Performance of Software by Detecting Fault

Abstract

Software systems in the modern computer world are notoriously complex and versatile. Finding newlineand fixing software design issues on a regular basis is thus essential. In the software newlinedevelopment process, bugs are the primary source of wasted time and money. Early failure newlineprediction increases system quality and dependability while decreasing software development newlinecosts. Software companies are increasingly engaging in the practice of fault-prone module newlineprediction before testing in order to spend resources effectively toward the creation of robust newlinesoftware. These fault prediction systems in software are only as good as the fault and associated newlinecode that was taken from prior software versions. Numerous studies have shown that software newlinemetrics are crucial components for predicting software errors. The field of software defect newlineprediction has also seen the development of several machine learning algorithms. It is crucial newlineto use machine learning methods to ascertain which group of metrics is most useful for fault newlineprediction. In our work, we provide a unique framework that integrates many state-of- the-art newlinemachine learning methods, including LGBM, XGBoost, Voting, AdaBoost, CatBoost, newlineGradientBoost, and Stacking. As per objectives listed in the synopsis, we have adopted the newlinemechanism to achieve these objectives. We studied various research proposals and identified newlinethe mechanism from concerned issues of literature review and then we adopted the approaches newlinethrough simulation tools and various existing methods like data collection, data visualization newlineetc. Our framework is shown in result chapter of thesis. We have also described interpretation newlineof our work as discussion and analysis. We have concluded with explored possibilities for newlinefuture research areas in specific domains.