Dynamic process model optimization method based on concept drift detection
Fan Zhang · 2024
The optimization of dynamic process model is always a difficult problem in process mining. Traditional model optimization methods assume that the model is stable, but most business changes with time in reality. The traditional static model optimization method cannot reflect the temporal change characteristics of the real process model. For this reason, this paper proposes an optimization approach of dynamic process model. Based on the detection and location of concept drift points, drift points to segment event logs precisely and extracts the optimal sub-model from the segmented logs by mining algorithm. The piecewise model composed of continuous optimal sub-models can effectively reflect the real state of the model at each stage. The proposed approach has got used in Tianyuan Big Data Transaction Platform, which shows that the optimization model has an advantage over the static model in fitness and precision