Cluster Analysis Of The Software Test Cases Based On The Community Structure

Software with uncertainty in the operation process, found in the operation of the error is likely to be in the testing phase failed to detect errors, so defects found in software operation, improve the reliability of software, reduce the loss caused by software defects and become particularly important. In this paper, the software implementation process as weighted complex networks, and aims to improve the efficiency of software testing, software testing cost.Based on complex network, software testing, and cluster analysis of related theories and research methods, analysis of the real open source software and build the weighted network divides the community to find the critical path to get test case and clustering, to reduce the test cases, to enhance the efficiency of the test.First, this paper proposes a weighted network software system modeling algorithm and software function in the operation process node, according to the function and the number of calls, call relations between the nodes add weight to the edge, to construct weighted network. Through improve the sparse matrix module increment, the maximum heap, and introducing auxiliary vector to improve the CNM algorithm, this paper proposes a weighted network community partition algorithm, define a module function Q, measurement network community classification effect.Secondly, this paper proposes a software test case generation algorithm based on the critical path, in all the communities from the root node traversal to leaf node, in the largest path midpoint strength sum is the critical path, the iteration to find the critical path to get the test case. Introduction of fuzzy set theory, this paper puts forward a kind of test case clustering algorithm, based on the generated test case structure similarity matrix, through the normalization and standardization, be partitioned matrix, the high similarity of cases classified as a class, in order to get the least amount of test cases.Finally, the community partition algorithm performance analysis, this article through gzip, tar, emacs program example, as a case study of tart program on to compare the different community partition algorithm prove that this algorithm has a superior place. Test cases for this article clustering algorithm performance analysis, seven files in Linux system as an example, and cyclomatic complexity calculation method to get the test cases and coverage of each file to compare, to prove the practicability of the algorithm. With real data prove that really save the test cases, improve the test efficiency.