A coincidental correctness test case identification framework with fuzzy C-means clustering

Cleansing coincidental correctness test cases has been proven to be useful in software fault localization. However, k-means clustering-based coincidental correctness test cases identification has not been studied yet. k-means clustering is hard classification and each sample point belongs to the cluster with the highest similarity, which leads to the inaccuracy of the cluster-based coincidental correctness. To address this issue, we propose an effective Coincidental Correctness test cases identification framework based on Fuzzy C-Means clustering (CC-FCM). The elements of coincidental correctness were first identified by probability function we designed, and the feature elements of the coincidental correctness were selected. Secondly, fuzzy c-means clustering was first introduced into identifying coincidental correctness test case after the dimensions of program execution traces were reduced. Finally, the results after coincidental correctness cleansing were used for the fault localization. To verify the effectiveness of the proposed CC-FCM, experiments were conducted by four fault localization methods, including Tarantula, Ochiai, Naish2 and Russel & Rao on 10 real-world subject programs. The experimental results showed that our proposed CC-FCM has a significant improvement over the compared methods, and that our approach has a lower false-positive rate and false-negative rate in coincidental correctness test case identification.