On the Use of Meta-Heuristic Algorithms for Automated Test Suite Generation in Software Testing

Abstract

There exists a dire need to automate the process of test suite generation to get the most optimal results as testing accounts for more than 40% of total cost. A solution consists of using meta-heuristic algorithms which iteratively improve the test data to reach the most optimized test suites. The goal of the study is to find the best suited algorithm to narrow down future research in the field of test automation and also provide issues on the design of new proposals. We focus on the performance evaluation of different major Meta-Heuristic Algorithms namely: Hill Climbing Algorithm (HCA), Particle Swarm Optimization (PSO), Firefly Algorithm (FA), Cuckoo Search Algorithm (CA), Bat Algorithm (BA) and Artificial Bee Colony Algorithm (ABC). Each algorithm is implemented to automatically generate test suites based on the program under test. Then, we develop a performance evaluation of each algorithm for five programs written in Java. The algorithms are compared using several process metrics (average time, best time, worst time) and also product metrics (path coverage & objective function values of the generated test suites). Results indicate ABC as the best suited algorithm as it gave the most optimal Test Suites in reasonable time. BA is the fastest one but produced less optimal results. FA is the slowest algorithm while CA, PSO and HCA perform in between. Some issues and strategies to create hybrid algorithms are discusses and pointed out.