A Method of Adaptive Process Mining Based on Time-Varying Sliding Window and Relation of Adjacent Event Dependency

Most existing process mining methods were designed for ignoring time variability from real business process data, thus it could be hard to implement adaptive process mining. To deal with this problem, a new method of adaptive process mining was proposed in order to mine unremittingly process models of gradual change which represents the improvement stages of business processes and improve accuracy of mined results. Given related concepts of a time-varying sliding window and relation of adjacent event dependency, update rules of modifying continuously size and progress in a time-varying sliding window were studied based on changed frequency of mined results and arrival rate of process instance streams, and an algorithm of process mining was presented by applying relation of adjacent event dependency among activities. Finally, a plug-in tool in PROM was developed to implement this algorithm.