| Software testing is an effective technique to guarantee software quality, and the pros and cons of test data decide test efficiency. Under the premise of no test result influenced, we can improve test efficiency and reduce the cost of test effectively by decreasing scale of test suite. Test data is the key of software testing. If main variables, that is principal input variables, of measured program can be analyzed and designing test data according to these variables, then we can improve the quality of test data and generating efficiency. Thus this thesis researches on analysis of main variables of program, and this work has significant theory meanings and broad application outlooks.This paper introduces the theory of principal component analysis (PCA) used in fault diagnosis and pattern recognition field to software testing, in order to improve test efficiency and reduce the cost of test. Accordingly, we present a program structure based main variable analysis of program method by using ant colony optimization (ACO) algorithm. The model of attribute reduction based on information entropy is used. When ACO algorithm is used to analyze main variables of program, the relationship between input/output variables, which is four arithmetic operations correlation and conditions judgement correlation, is used as important parameters—the heuristic information. Redundant Strapped-Down Inertial Measurement Unit (RSDIMU) is chosen to be the measured program, and the method we present is used to analyze main variables of RSDIMU. Experimental results show that the approach is promising and effective, and main variables of program can be determined more accurately by using the two correlations to calculate heuristic information. Furthermore, test data designed according to main variables can improve test efficiency and reduce the cost of test effectively. |
