Genetic Algorithm Based on Software Diagnosis Testing

Based on genetic algorithm, the problem of software diagnosis testing is considered in this paper to reproduce the failure for the kind of systems with multi-input/output variables. Firstly, the problem prototype is abstracted, and then solutions to the prototype problems are introduced. From the aspects of coding scheme, population initialization, genetic operation, selection of fitness function, and convergence criterion, etc., how genetic algorithm can be applied in such prototype problem as software fault reproduction is thoroughly described. The experimental result shows that the injected software failure can be reproduced rapidly in the given program.