The Study Of Multi-path Regression Test Data Generation Method With Multi-population Genetic Algorithm

In recent years,the researches on the method of generatig regression test data through some useful information of the existing test data have received widely attention.But most of the research findings in this area are suitabe for just a single generating path.Only quite few are multi-path oriented,and there is a long way to go.Genetic algorithm is often used to solve the problem of multi-path coverage,and currently,genetic algorithm is used to generate multipath regression test data,mainly using multiple paths to serial,thereby the efficiency of the regression test is not high.For this reason,this article puts forward a new multi-path oriented method of generating regression test data based on multi-population genetic algorithm.The new method takes full advantages of the useful information of the existing test data,and generates test data covering all target paths through performing just one genetic algorithm.The main tasks of this research include:(1)To analyse and study the basic theories and methods of regression testing,path testing and genetic algorithm.So that the theoretical basis of generating test data method which covers multiple path can be provided.(2)To build up a multi-objective optimization model which can deal with most of the problems of multi path test data generation,and adopt a improved multi-population genetic algorithm to solve the proposed model.To avoid local convergence of traditional genetic algorithm,in this paper,we improve structure of the population,genetic operators and migration strategies of multi-population genetic algorithm.In addition,the number of sub-populations has a great influence on the efficiency of the algorithm.In order to avoid excessive number of sub-populations resulting in the increase of load of the algorithm.The reduction and grouping of target paths are needed.The paper use unreachable path detection model to find unreachable paths and then remove unreachable paths which reduce the scale of target paths.And then grouping the target paths by the path similarity,each group corresponds to a sub-population,thereby reducing the number of sub – populations.(3)To design an individual evaluation function which uses genetic algorithm to solve multi-objective optimization model.The fitness function's construction is based on the thought of path similarity,and the factor of matching node location is also considered in the function.The node weights are added according to the sequence of the nodes matched by target path and actual path executed by test data.The position of matching node is more forward,the greater the weight,and thus the fitness value of the individual is greater.(4)To develop a multi-path regression test data generation system with java on the Eclipse platform.The system integrates the method of this article and the existing regression test data method,and then run the program to be tested by these methods.It is verified that the method of this article can improve the efficiency of regression testing by comparing the operating results of the system.