A Software Impact Analysis Tool based on Change History Learning and its Evaluation

Abstract

Software change impact analysis plays an important role in controlling software evolution in the maintenance of continuous software development. We developed a tool for change impact analysis, which machine-learns change histories and directly outputs candidates of the components to be modified for a change request. We applied the tool to real project data to evaluate it with two metrics: coverage range ratio and accuracy in the coverage range. The results show that it works well for software projects having many change histories for one source code base.