Measuring Software Change Impact Using Requirement Dependencies

Tracking software changes and their impact are important tasks when developing and deploying software. Software developers armed with the knowledge of which changes have the highest impact on the existing software can selectively deploy changes to minimize any negative impacts on the deployment. Software developers can use requirements change impact measures to assist with how to implement the changes to software code. One challenge when identifying change impact is quantifying the effect of the impact. In this research, we approach the problem of identifying and quantifying the impact of changes by applying IR techniques along with data mining. We apply TF-IDF and LSI to help determine which changes have a high potential of modifying existing requirements. We also analyze similarities among requirements changes and similarities among requirements baselines. In both cases, our threshold for identifying similarity is between 75% - 95%. We use these transitive relationships in conjunction with the change to baseline relationships to derive more insight about the inter-relationships of the baseline requirements and changes to those requirements. We bucket these relationships according to their impact scores into buckets of low, medium, and high impact using per dataset analysis of tertiles.