Understanding Coding Behavior: An Incremental Process Mining Approach

Capturing and analyzing interaction data in real-time from development environments can help in understanding how programmers handle coding activities. We propose the use of process mining to learn coding behavior from event logs captured from a customized Integrated Development Environment, concerning interactions with both such an environment and a Version Control System. In particular, by using an incremental approach, the discovered model can be refined after every single development session, which avoids the need to for the model to learn from scratch from previous sessions. It would also allow one to provide the programmer timely suggestions to improve their performance. In this paper, we applied off-line incremental behavior, so as to be able to analyze it at several levels of depth and at different moments. As a preliminary evaluation of our approach, we investigated the coding activities of six novice students of a Java academic programming course working on a programming case study. The results provide some useful information about the initial difficulties in coding activities faced by programmers and show that their coding behavior could be considered as “formed” after a development task requiring approximately 4000 rows of code.