Defect prediction for Cascading Style Sheets

Abstract

Testing is a crucial activity in software development. However exhaustive testing of a given software is impossible in practice because projects have serious time and budget limitations. Therefore, software testing teams need guidance about which modules they should focus on. Defect prediction techniques are useful for this situation because they let testers to identify and focus on defect prone parts of software. These techniques are essential for software teams, because they help teams to efficiently allocate their precious resources in testing phase. Software defect prediction has been an active research area in recent years. Researchers in this field have been using different types of metrics in their prediction models. However, value of extracting static code metrics for style sheet languages has been ignored until now. User experience is a very important part of web applications and its mostly provided using Cascading Style Sheets (CSS). In this research, our aim is to improve defect prediction performance for web applications by utilizing metrics generated from CSS code. We generated datasets from four open source web applications to conduct our experiments. Defect prediction is then performed using three different well-known machine learning algorithms. The results revealed that static code metrics based defect prediction techniques can be performed effectively to improve quality of CSS code in web applications. Therefore we recommend utilizing domain-specific characteristics of applications in defect prediction as they result in significantly high prediction performance with low costs. (C) 2016 Elsevier B.V. All rights reserved.