.png){kind=icon}
.png){kind=icon}
Return
Process mining and path similarity analysis
.png){kind=icon}
PRE
AI
DOI：10.1016/j.iswa.2026.200660.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This article proposes a method to leverage process mining techniques to analyse real-life event logs and give insights to managers to improve process performance. For this purpose, we adopt the Levenshtein distance and k-medoids clustering, to identify representative prototype traces for process variants. Then, we introduce a novel Composite Similarity Score, integrating graph-based and attribute-based measures, to assess trace conformance to prototypes. Thus, anomalous traces can be identified by means of outlier detection, revealing significant deviations in duration and process complexity. Key findings highlight prolonged durations and bottlenecks, suggesting targeted process optimization opportunities. By deriving a standardized to-be process model from prototypes, we face process standardization, to allow institutions and enterprises to enhance efficiency, reduce cancellations, and improve decision-making. We test this method on loan application BPI Challenge 2017 dataset.
Keywords:
Path similarity
K-medoids clustering
Mahalanobis distance
Outlier detection
Process standardization
Loan application process
Journal
I
IF:
4.3
Papers:
90
Citations:
0