Development and Research of the Test Data Generation Approach Modifications

In this paper, we investigate approaches to solving the problem of generating data for a set of source code paths using a genetic algorithm. In the first approach, a genetic algorithm is used to find the most adapted chromosome, which is a set of test data that provides passage along the most difficult path. The set of data sets that provide maximum code coverage is found by iterative repetition of running the genetic algorithm with preliminary zeroing of the weights of operations corresponding to previously found chromosomes. Another approach assumes that the fitness function initially includes two components. The first component, as in the first approach, is responsible for the complexity of each path. The second component is responsible for the maximum possible differences in population pathways. For the second approach, modifications of the algorithm were developed, which are aimed at solving the problem of unstable convergence of the approach.