Identifying class integration test order using an improved genetic algorithm-based approach

Abstract

Software testing is a very difficult activity, representing a large part from a software system’s development process. Within the class-based integration testing methodology, determining the order in which the application classes have to be tested is of major importance for reducing the testing time and cost. The Class Integration Test Order (CITO) problem deals with identifying the testing order of classes which minimizes stub creation effort, and subsequently testing cost. In this paper we propose an efficient approach using a genetic algorithm with stochastic acceptance for determining the class integration test order which minimizes the stubbing effort needed during the class-based integration testing. In our proposal, we estimate the stub creation complexity by allocating weights to different types of dependencies between the classes in the software system’s Object Relation Diagram. Four synthetic examples and six software systems often used in the CITO literature are used as case studies for experimentally evaluating our proposal. The effectiveness of our approach is confirmed by the obtained results that outperform the existing related work which provide experimental results on the case studies considered in this paper.