Bug Severity Classification in Issue Tracking Systems with Support Vector Machines

Abstract

Bug severity classification helps software maintenance teams identify which issue reports require urgent review, faster assignment, or deeper technical investigation. This article presents a Support Vector Machine-based framework for classifying bug severity from issue tracking data using report text and selected metadata. The framework cleans bug summaries and descriptions, preserves important technical tokens, extracts TF-IDF text features, encodes component and report metadata, prepares severity labels, and evaluates classwise prediction behavior. The study shows that SVM classification can separate low and critical severity reports more clearly than medium and high severity reports, where wording and impact boundaries are often less distinct. The proposed approach supports triage teams by ranking likely severe defects, reducing manual screening effort, and identifying ambiguous reports that still require expert review.