A Dead Code Evaluation Method based on Complex Network

Abstract

Dead code is widespread in open-source and commercial software systems. While there is some work on dead code detection, there is no work on evaluating the risk of removing dead code. This paper introduces complex network into dead code evaluation and proposes a dead code evaluation method based on weighted technique for order preference by similarity to ideal solution(TOPSIS). We regard degree centrality, closeness centrality, betweeness centrality and the proportion of alive codes as the multi-attribute of weighted TOPSIS, which overcomes the shortage of using the same weight for each attribute in the original method. We evaluate dead code in two open-source Java projects. The results show that this method can well evaluate the risk of deleting dead code in nodes.