A method for analyzing fault sequence feature based on clustering

A large number of fault sequences emerge while analyzing software security,so mining fault sequences is significant for determining the locations ofsoftware faults. In this paper, clustering technology is adopted to analyzefault sequences. In order to cluster unequal-length fault sequences, we proposea new similarity measure and develop a sequence clustering algorithm called CSE(Clustering based on Sequence Elements). In CSE, the number of the mutualsequence elements contained in sequences is calculated to determine thesimilarity of sequences, and fault sequences are collected into clustersaccording to the similarity measure. To analyze the feature of software faults,we introduce a clustering-based method, which combines clustering analysis andsequence alignment. Experimental results show that CSE has higher clusteringquality and good scalability. ICIC International