A New Spectrum-based Fault Localization Method by Using Clustering Algorithm

Abstract

Locating the errors or faults in the software debugging process is very expensive and time-consuming. Therefore, automated debugging has become popular due to its capability to reduce cost and increase the efficiency. Spectrum-based method is one of the most important automated fault localization methods in the Software debugging techniques. Most of the existing spectrum-based fault localization methods against some others, whereas the comparison of presented methods is not comprehensive due to the different test case set is utilized. This paper implemented the existing 28 methods to evaluate the spectrumbased fault localization methods in the same test case set. Furthermore, we proposed a new spectrumbased fault localization method, which utilize k-means algorithm, to obtain a suspicious ranking of program statements and capture the statement with accepted high suspiciousness, so as to improve the effectiveness of fault-localization. An experiment is designed and we apply the proposed method to our java plug-in project of spectrum-based fault localization. The experimental results show that this method can locate faults in the software effectively and has good performance.