Applying Semantic Analysis to Feature Execution Traces

Recently there has been a revival of interest in feature analysis of software systems. Approaches to feature location have used a wide range of techniques such as dynamic analysis, static analysis, information retrieval and formal concept analysis. In this paper we introduce a novel approach to analyze the execution traces of features using Latent Semantic Indexing (LSI). Our goal is twofold. On the one hand we detect similarities between features based on the content of their traces, and on the other hand we categorize classes based on the frequency of the outgoing invocations involved in the traces. We apply our approach on two case studies and we discuss its benefits and drawbacks.