Finding suspicious patterns of object-oriented programs based on variance analysis

Several variance analysis methods that are used in fault localization based on program behaviors were analyzed and compared. None of these methods takes into account the features of object-oriented programs, thus, could not be used accurately in fault localization in object-oriented programs. The present study proposed that when doing variance analysis, object behaviors could be described by using the object life behavior model; then a behavior model of the object-oriented program was constructed based on the model; the similarity of program behaviors was described by using the sequence pattern. Based on the variance analysis of program behaviors, the method of finding suspicious patterns of object-oriented programs was proposed. The model of finding suspicious patterns was also constructed. This model consisted of preparing data, extracting patterns, and finding suspicious patterns. The standard for pattern extracting was defined and the mining algorithms of finding patterns were also described. At last, the method of finding suspicious patterns based on variance analysis and its implementation procedure were depicted in detail. This method has been used in the research of fault localization and has been very effective. This work is supported by the National Natural Science Foundation of China under grant No. 60603039.