A Profile Clustering Based Event Logs Repairing Approach for Process Mining

Abstract

Process discovery, as the most crucial learning task in the process mining, builds some highly complex process models such as “spaghetti-like” from event logs contained large amounts of data. To enhance the process discovery method for all of flexible environments, many researchers tried to exploit trace clustering approaches to split the logs into several homogeneous sub-logs, which are used to generate the corresponding sub-process model, respectively. However, their works are based on the assumption that the event logs are complete without missing any data values. On the contrary, the data in an event log may be lost due to some reasons such as system failure and human error. In this paper, we propose a method to deal with incomplete logs so as to discover the process model. First, we split up the event logs based on trace clustering. Then, the missing traces are assigned to the most similar clustering results, respectively. After that, with supplementing the missing data in the trace, a corresponding sub-process model is mined using the proposed method. At last, some experimental results on three real-life complex event logs demonstrate the feasibility and effectiveness of our method.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 30 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.