APPROACH FOR MINING SOFTWARE EVOLUTIONARY COMMUNITY OUTLIERS BASED ON COMMUNITY-MATCHING

摘要

Study on detecting a tiny fraction of influential nodes in software evolution is very significant for predicting software updating trends, facilitating software development and version refactoring. We exploit recent advances in mining functions with obvious change to better understand the feature of software evolution. In this paper, detecting software evolutionary community outliers is our main work. Firstly, algorithm Depth-First-Search Weight (DFS- Weight) is proposed to construct the Weighted Function Dependency Network (WFDN). Secondly, we use key-nodes based approach Function Be-longingness Matrix Generating (FBM-Gen) to detect the community structure of WFDN, and then find the probability distribution of function nodes in each community. Thirdly, community-matching based algorithm Software Evolutionary Community Outliers Detection (SECO-Detection) is put forward. It generates software evolutionary community outliers which evolve in a different way relative to other community members. Finally, experimental results on both real and synthetic datasets show that the proposed approach is highly effective in discovering interesting evolutionary community outliers.