Cluster-based adaptive test case prioritization
Xiaolin Wang,Sulan Zhang
DOI: https://doi.org/10.1016/j.infsof.2023.107339
IF: 3.9
2023-10-12
Information and Software Technology
Abstract:In order to enhance the efficiency of regression testing, test case prioritization (TCP) has been widely implemented, wherein a higher priority test case is executed earlier. Traditional TCP methods focus on improving the prioritization algorithm's efficacy. However, the majority of TCP approaches are characterized by a predetermined sequence of test cases prior to execution. Once established, this sequence remains consistent throughout the entire test execution process. As a result, any execution information generated during current test execution (such as fault-detected information) is unavailable for use in current round of test case prioritization and can only be utilized in subsequent regression testing. To address the issue of lagging utilization of fault-detected information, a cluster-based adaptive test case prioritization approach is proposed, which adds the new adaptive adjustment content in pre-prioritization. First, a new clustering criterion is defined and designed, by which produces test-case clusters in advance. Second, an adaptive TCP algorithm is proposed, which utilizes fault-detected information to adaptively adjust the order of test cases during the execution process based on the test-case clusters. Finally, one open-source Java program and three industrial-grade Java programs were selected for empirical evaluation. The experimental results demonstrate that the proposed technique not only serves as an enhanced version of pre-prioritization to improve the performance of the corresponding pre-prioritization technique, but also functions as an independent approach that outperforms other TCP techniques, including cluster-based TCPs, and another adaptive TCP. Specifically, when step=2 is applied using our cluster-based adaptive TCP approach, the results are significantly better than those obtained with step=1. For instance, in CT-14 , the median APFD improvement rate for step=2 reaches 17.08 %, which is substantially higher than that achieved with step=1 (5.48 %).
computer science, information systems, software engineering