Improving severity prediction on software bug reports using quality indicators

Recently, research has been conducted to explore the prediction schemes to identify the severity of bug reports. Several text mining approaches have been proposed to facilitate severity prediction. However, these studies mainly focus on the textual information of the bug reports. Other attributes of the bug reports have not been comprehensively discussed. In this paper, we investigate the influences of four quality indicators of bug reports in severity prediction. In an empirical study with the Eclipse dataset, the results show that considering these indicators can further improve the performance of a previous work employing only textual information.