Loading...

Discovering Business Processes of an Organization Using Process Mining Approach and Implementing Prediction Models Based on Machine Learning

Fatehi, Mahshid | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57161 (01)
  4. University: Sharif University of Technology
  5. Department: Industrial Engineering
  6. Advisor(s): Rafiee, Majid
  7. Abstract:
  8. In today's competitive world, organizations must optimize their processes and minimize costly deviations to ensure survival in the market. The occurrence of unknown output and the potential for non-conformance during ongoing cases often lead to reactive procedures and subsequent challenges. In this situation, Predictive Process Monitoring emerges as a valuable solution to address this issue. This article focuses on predicting non-conformance cases within an Iranian fast-moving consumer goods company. Initially, the process model is simplified through trace clustering methods. Two encoding techniques, TF-IDF and one-hot, are employed to preprocess the event log before applying four selected clustering methods. Comparative analysis reveals that pairing the mini-batch K-means algorithm and TF-IDF encoding method outperforms other combinations across multiple metrics. In the predictive monitoring stage, we introduce a novel hybrid encoding method, combining log-skeleton and aggregation algorithms. The encoded data is then fed into the XGBoost classification algorithm for prediction generation. The proposed framework demonstrates exceptional prediction results with F1-score and AUC-ROC metrics above 0.8 and 0.9 respectively in our real-life case, which is studied in the process mining field for the first time. Ultimately, these prediction outcomes yield valuable managerial insights
  9. Keywords:
  10. Process Mining ; Predictive Process Monitoring ; Machine Learning ; Data Encoding ; Trace Clustering ; Predictive Model

 Digital Object List

 Bookmark

No TOC