COMPARISON OF EVOLUTIONARY ALGORITHMS FOR TEST DATA GENERATION

Abstract

Software testing is a very important process and plays a very important and key role in software industry. The cost of testing consumes a significant portion of the total project cost. Exhaustive testing is not possible and hence testing the focus is on testing those portions of the project or program where probability of finding fault is maximum. Test Data Generation is an important part of testing and it is the process of creating a data set which is then applied on the new or revised software for testing it. Now a day’s focus has shifted on automatic generation of test data which saves both time and effort. In the light of generating test data automatically various methods and algorithms are being developed and used so that the problem of generating test data can be solved not only efficiently but also in less time. In this thesis we have compared three important Test Data Generation Algorithms namely Ant Colony Optimization (ACO), Artificial Bee Colony (ABC) and Genetic Algorithm (GA). These algorithms generate test path from CFG of the program and corresponding to that test data is generated that satisfies that path. We have developed a tool named “TEST GENERATOR COMPARATOR” that takes input, CFG of 10 C programs and then the tool applies these three algorithms on each program CFG. Test data is generated by each algorithm for each input program, and for each algorithm and each program the tool outputs three parameters namely number of iterations, path coverage and the time taken. Based on these three parameters the algorithms are compared. The results obtained show that ABC gives better result as compared to other two algorithms and hence is well suited for test data generation.