Bug Severity Prediction Using a Hierarchical One-vs.-Remainder Approach

摘要: Assigning severity level to reported bugs is a critical part of software maintenance ensure an efficient resolution process. In many bug trackers, e.g. Bugzilla, this time consuming process, because reporters must manually assign one seven levels each bug. addition, some types may be more often than others, leading disproportionate distribution labels. Machine learning techniques can used predict the label newly automatically. However, from imbalanced data in multi-class task remains major difficulties for machine classifiers. paper, we propose hierarchical classification approach that exploits class imbalance training data, reduce bias. Specifically, designed tree consists multiple binary classifiers organised hierarchically, such instances most dominant are trained against remaining classes but not next tree. We FastText classifier test and compare between standard approaches. Based on 93,051 reports 38 Eclipse open-source products, was shown perform relatively well with \(65\%\) Micro F-Score \(45\%\) Macro F-Score.