A Stream Data Mining Approach to Handle Concept Drifts in Process Discovery

Process discovery algorithms discover process models from event logs recorded from the real-life processes. Traditional process discovery algorithms assume that logged processes remain in a steady state over time. However, this is often not the real-world case due to concept drifts. To continue using well-defined, off-line process discovery algorithms to process a stream of process execution traces, we propose an online approach that performs concept drift detection and adaption of process models discovered with traditional process discovery algorithms. Experimental results explore the effectiveness of the proposed approach coupled with several traditional process discovery algorithms.