Research On Fuzzy Testing Method Based On Multi Population Genetic Algorithm

With the scale of the software project getting bigger and bigger,finishing a high quality software project not only depends on software development process,but also depends on the testing process.In the process of software testing,test cases generation is the key factors of identifying defects in a software system.As a result,the generation of test cases in software testing plays a key role in software testing.At the present stage,the generation of software test cases based on genetic algorithm has been widely concerned and studied.The traditional genetic algorithm has some blindness in the data search,and it also has some disadvantages such as "premature convergence" and slow convergence speed in the process of evolution.Therefore,the point of the research is how to improve the traditional genetic algorithm.This paper analyzed the methods of traditional genetic algorithm to guide the generation of the software test cases,and we find out that traditional genetic algorithm has some blindness in the data search,and it also has some disadvantages such as "premature convergence" and slow convergence speed in the process of evolution.This paper proposes a method based on multi population genetic algorithm as the guidance of Fuzzy test case generation.We static analyze and dynamic analyze the binary file of the software program to obtain genetic algorithm adaptation degree function parameters,and calculate the fitness of each individual in a multi population genetic algorithm;we divid different populations according to the fitness and Hamming distance of each individual;the different populations evolute and interspecific hybridizate alternately;we eventually transport the individual which has high fitness to the output population.In order to speed up the convergence rate of the algorithm,we use the strategy of the elite in the process of output population evolution and the output population is the fuzzing test cases set.The static analysis and dynamic taint analysis lead to the data generation of the genetic algorithm based on critical path which can effectively reduces the blindness of searching data in traditional genetic algorithm.Besides,the hybridization of multi population avoids the premature phenomenon in the traditional genetic algorithm and the output of the population can effectively improve the convergence speed of traditional genetic algorithm.This paper builds a test cases generation tool based on the multi population genetic algorithm and this tool includes three models:static analysis module,dynamic analysis module and test case generator.We use this tool on three basic procedures and MATLAB functions to verify the advantages of our algorithm.The results show that multi population genetic algorithm is better than traditional genetic algorithm in evolutionary generation and time.