Variable Dependency Graph Summarization

Abstract

In personalized software, collected user data is used to give a tailored user experience. A user might be interested in understanding how their data (inputs) resulted in their personalized output. The Variable Dependency Graph (VDG) can explain how inputs of a program flow to the output. However, with increasing program size, there is a need for summarizing the VDGs and understanding these summarizations.

The first contribution of this thesis is the exploration of a fitting technique to summarize the VDG for a user. The result of the search was the technique TG-SUM, which supports all the VDGs graph characteristics that this thesis identified, as well as facilitates the visualization for a user. The second contribution of this thesis was to help a user interpret the VDG summarization by experimenting with how different code patterns tend to be summarized.