Process-Aware Enterprise Social Network Prediction and Experiment Using LSTM Neural Network Models

• Main Authors: Dinh-Lam Pham, Hyun Ahn, Kyoung-Sook Kim, Kwanghoon Pio Kim
• Format: Article
• Language: English
• Published: IEEE 2021-01-01
• Series: IEEE Access
• Subjects: Process mining; long short-term memory neural network; process-aware enterprise social network; next event prediction
• Online Access: https://ieeexplore.ieee.org/document/9399143/

Process mining that exploits system event logs provides significant information regarding operating events in an organization. By discovering process models and analyzing social network metrics created throughout the operation of the information system, we can better understand the roles of performers and characteristics of activities, and more easily predict what will occur in the next operation of a system. By using accurate and valuable predicted information, we can create effective environments, provide suitable materials to perform activities better, and facilitate more efficient operations. In this study, we apply the long short-term memory, a variant of the recurrent neural network, to predict the enterprise social networks that are formed through information regarding a business system’s operation. More precisely, we apply the multivariate multi-step long short-term memory model to predict not only the next activity and next performer, but also all the variants of a process-aware enterprise social network based on the next performer predictions using a probability threshold. Furthermore, we conduct an experimental evaluation on the real-life event logs and compare our results with some related researches. The results indicate that our approach creates a useful model to predict an enterprise social network and provides metrics to improve the operation of an information system based on the predicted information.