Process Mining Applied to the BPI Challenge 2012: Divide and Conquer While Discerning Resources

Abstract

A real-life event log, taken from a Dutch financial institute, is analyzed using state-of-the-art process mining techniques. The log contains events related to loan/overdraft applications of customers. We propose a hierarchical decomposition of the log into homogenous subsets of cases based on characteristics such as the final decision, offer, and suspicion of fraud. These subsets are used to uncover interesting insights. The event log in its entirety and the homogeneous subsets are analyzed using various process mining techniques. More specifically, we analyze the event log (a) on the resource perspective and the influence of resources on execution/turnaround times of activities, (b) on the control-flow perspective, and (c) for process diagnostics. A dedicated ProM plug-in developed for this challenge allows for a comprehensive analysis of the resource perspective. For the analysis of control-flow and process diagnostics, we use recent, but pre-existing, ProM plug-ins. As the evaluation shows, our mix of techniques is able to uncover many interesting findings and could be used to improve the underlying loan/overdraft application handling process.