Test cases generation for multiple paths based on PSO algorithm with metamorphic relations

Abstract

The generation of multiple-path test cases can greatly enhance the efficiency of path-wise testing. Various methods adopting meta-heuristic algorithm to generate multiple-path test cases have been proposed, but existing methods focus on improving the meta-heuristic algorithm to get better test case generation efficiency, and test cases covering each path needs to be generated by meta-heuristic algorithm searching. To improve efficiency, a test case generation method for multiple-path coverage is proposed in this study, which combines a particle swarm optimisation (PSO) algorithm with metamorphic relations (MRs). The method first generates a test case using the PSO algorithm, and then generates new test cases by repeatedly using MRs between test cases. This method reduces evolving numbers of PSO algorithm. The experimental results show that the proposed method can significantly enhance the efficiency in terms of fitness evaluations and time consumption.

Crow Search Algorithm with Improved Objective Function for Test Case Generation and Optimization

“…The FF combines approach level and branch distance in GA for branch coverage. Lv XW et al [16] proposed the PSO approach with metamorphic relationships for generating test cases for multiple path coverage. The aim was to cover multiple paths efficiently with fewer iterations.…”

“…In this work, we consider two case studies for test cases by covering the multiple paths. The first one is a python program for Triangle Identification Problem [16,18] and generates test cases with 100% path coverage. The second example is a small python program to Identify Armstrong Numbers.…”

“…The BD determines the degree to which the input deviates from the predicate [16]. It specifies whether the test data is close or far to satisfy the conditional statement of the code [25].…”

“…Path coverage, statement coverage, decision coverage, and branch coverage are all frequently used criteria [15]. For covering the target path, many researchers [10,16] employ a FF based on branch distance. Branch distance is estimated by examining the branching node's conditional statement and sometimes missed target [17].…”

“…The branch distance is paired with heuristic information of approximation level to improve the searching. It indicates the distance between the target node and the node currently visited [16]. When there is a low probability of covering a path, this combined FF fails [18].…”