An xgboost-based model for predicting nature-based bug reports

Abstract

Because bug reports (BRs), which contain comprehensive information such as the description, status, a reporter, assignee, priority, severity, and other details about the bug, are a valuable tool for fixing defects found during software testing, researchers have focused their attention on the routine upkeep of software development systems. Since there are a lot more BRs in this process than there are problems in the system, the primary challenge is analyzing these BRs to find every issue. This is a laborious and time-consuming effort to conduct by hand. Thus, the optimal option is the automated one. Much of the current research concentrates on automating this procedure from many angles, such figuring out how serious or urgent the defect is. They did not, however, take into account the fact that the defect is a multi-class categorization issue. This research provides a solution to this issue by putting forward a novel prediction model to examine BRs and forecast the kind of bug. The suggested model combines machine learning and NLP (natural language processing) methods to build a combined machine learning algorithm. Using a publicly accessible dataset for three online software bug repository (Mozilla and Eclipse), that comprises six types (Program Anomaly, GUI, Networking or Security, Configuration, Efficiency, and Test-Code), we mimic the suggested model. The results of the simulation demonstrate that the suggested model can outperform the majority of current models in terms of accuracy, achieving 90.42% without text supplementation and 96.72% with it.