Evolutionary Software Test Data Generation Using Sequence Comparison

Evolutionary algorithms have been applied to the dynamic structural test data generation problem. The fitness evaluation methods proposed so far either suffer from poor guidance characteristics or can not be applied uniformly. We propose a new method based on the pairwise sequence comparison techniques defined in bioinformatics. Interpreting the desired and generated paths as sequences, aligning and comparing these paths gives an intuitive measure about their similarity. We also define a sequence distance which measures the amount of change needed to convert the generated path into the desired path. The similarity and distance scores can then be used for evaluating candidate solutions in a search-based test data generation problem. Our preliminary tests show that this method produces promising results.