Weighted string distance approach based on modified clustering technique for optimizing test case prioritization

Abstract

Numerous test case prioritization (TCP) approaches have been introduced to enhance the test viability in software testing activity with the goal to maximize early average percentage fault detection (APFD). String based approach had shown that applying a single string distance-based metric to differentiate the test cases can improve the APFD and coverage rate (CR) results. However, to precisely differentiate the test cases in regression testing, the string approach still requires an enhancement as it lacks priority criteria. Therefore, a study on how to effectively cluster and prioritize test cases through string-based approach is conducted. To counter the string distances problem, weighted string distances is introduced. A further enhancement was made by tuning the weighted string metric with K-Means clustering and prioritization using Firefly Algorithm (FA) technique for the TCP approach to become more flexible in manipulating available information. Then, the combination of the weighted string distances along with clustering and prioritization is executed under the designed process for a new weighted string distances-based approach for complete evaluation. The experimental results show that all the weighted string distances obtained better results compared to its single string metric with average APFD values 95.73% and CR values 61.80% in cstcas Siemen dataset. As for the proposed weighted string distances approach with clustering techniques for regression testing, the combination obtained better results and flexibility than the conventional string approach. In addition, the proposed approach also passed statistical assessment by obtaining p-value higher than 0.05 in Shapiro-Wilk’s normality test and p-value lower than 0.05 in Tukey Kramer Post Hoc tests. In conclusion, the proposed weighted string distances approach improves the overall score of APFD and CE and provides flexibility in the TCP approach for regression testing environment.