Predicting Defect Types in Software Projects

Predicting software defects has been one of the most demanding tasks for software engineering researchers and practitioners. The work in this area resulted in producing various defect prediction models. Their common weakness is that they typically treat all defects equally. However, software companies need to categorize defects found in their products to estimate user satisfaction and to prioritize which defects are supposed to be fixed first. Several classifications are supported by popular defect tracking systems, such as Bugzilla. The Defect Types Model introduced in this paper estimates proportions of defects of various types categorized by their severity for users. This model contains two groups of factors: controllable (process quality) and uncontrollable (features of software). It incorporates results from statistical analysis of the ISBSG project dataset adjusted by other reported results and expert knowledge. This model can be used either as a standalone model (predicting proportions of defects) or in combination with other defect prediction model (predicting number of defects of different types).