Entropy-Based Behavioral Closeness Filtering Chaotic Activity Method

Abstract

In the era of big data, one of the key challenges is to discover process models and gain insights into business processes by analyzing event data recorded in information systems. However, Chaotic activity or infrequent behaviors often appear in actual event logs. Process models containing such behaviors are complex, difficult to understand, and hide the relevant key behaviors of the underlying processes. Established studies have generally achieved chaotic activity filtering by filtering infrequent activities or activities with high entropy values and ignoring the behavioral relationships that exist between activities, resulting in effective low-frequency behaviors being filtered. To solve this problem, this paper proposes an entropy-based behavioral closeness filtering of chaotic activities method. Firstly, based on the behavior profile theory of high-frequency logging activities, the process model is constructed by combining the feature network and the module network. Then, the identification of suspected chaotic activity sets is achieved through the Laplace entropy value. Next, a query model is built based on logs containing suspicious chaotic activity. Finally, based on the succession relationship, the behavioral closeness of the query model and the business process model is analyzed to achieve the goal of accurately filtering chaotic activities to retain behaviors beneficial to the process. To evaluate the performance of the method, we validated the effectiveness of the proposed algorithm in synthetic logs and real logs, respectively. Experimental results showed that the proposed method performs better in precision after filtering chaotic activities.