Using testing trace for automatic user categorization

Abstract

Testing has always been an indispensable part of software development. With the increasing amount of testing, the volume of data and information generated from testing grows substantially. The question arises on how to take advantage of the testing data, besides traditional coverage and debugging. In this paper, we propose an approach of using test trace data of a software application to its run-time user categorization. It collects test execution trace of programs studied by the software tool, and derives internal metrics of different categories from the trace information. During run time, we look at the user's artifacts as well as the user's behavior to categorize them into predetermined groups and serve them accordingly. Our work in-progress is to apply this method to a software product line, PolyFlow, including a web service that generates, runs, and analyzes test cases of programs under study. One benefit of our method is that it does not require storage of user profiles.