Mining Change Patterns in AspectJ Software Evolution

Understanding software change patterns during evolution is important for researchers concerned with alleviating change impacts. It can provide insight to understand the software evolution, predict future changes, and develop new refactoring algorithms. However, most of the current research focus on the procedural programs like C, or object-oriented programs like Java; seldom effort has been made for aspect-oriented software. In this paper, we propose an approach for mining change patterns in AspectJ software evolution. Our approach first decomposes the software changes into a set of atomic change representations, then employs the apriori data mining algorithm to generate the most frequent itemsets. The patterns we found reveal multiple properties of software changes, including their kind, frequency, and correlation with other changes. In our empirical evaluation on several non-trivial AspectJ benchmarks, we demonstrate that those change patterns can be used as measurement aid and fault predication for AspectJ software evolution analysis.