A. Watkins (USA)
Genetic Algorithms, Software Test Data Generation, Path Testing.
Software testing is a field of paramount importance as billions of dollars are lost each year through faulty software. Testing programs requires test data, and genetic algorithms have been used automatically to generate this test data. While all these compare their generation technique to random generation, most have not compared it to others that also use genetic algorithms. This paper looks at two such methods for test data generation aand runs tests on two benchmark programs to determine if one fitness function is better than another. The results are then compared to a third fitness function which should be the worst case performance for the genetic algorithm. The conclusions indicate that the two fitness functions from the literature perform slightly better than the worst case fitness function. It is more difficult however, to determine which of the two will perform the best in all situations and this is an area for future research.
Important Links:
Go Back
