RP2A: Rare Process-Pattern Analysis - Identifying Potential Problem Process-Patterns by Analyzing System Execution Data

• Published in: Proceedings / Asia Pacific Software Engineering Conference pp. 432 - 436
• Main Authors: Saito, Shinobu, Takahashi, Kenji, Hamada, Yasuyuki, Vermeulen, Jan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2022
• Subjects: Business; Business processes; Companies; event log; Industries; rare patterns; Search problems; Software engineering; Systems operation
• ISSN: 2640-0715
• DOI: 10.1109/APSEC57359.2022.00057

As companies run their business (e.g., sales, pro-curement), they conduct a set of business processes that include activities and flows connecting them. There exist an enormous number of patterns of business processes, or process-patterns. This industrial paper proposes a new technique, RP2A (Rare Process-Pattern Analysis), to examine process-patterns for finding operational problems such as compliance violations. The technique analyzes system execution data (= event log and user information) stored in the system related to the processes. By a combination of process-pattern search and activity search, RP2A discovers potential problem process-patterns in terms of execution frequency. We also report an industrial case study carried out using the technique. By analyzing the system execution data corrected from the 6-month system operation, we identified all 373 process-patterns, and 32 potential problem process-patterns out of them. Then, 4 domain experts in the industry verified the 32 process-patterns and found that 18 process-patterns, namely more than half of them (56% (=18/32)) contain actual problems. Almost problems are violations of internal policies which even the domain experts did not recognize. The technique succeeded in finding a kind of unknown problems in the business processes.