dinhlampham / predictiveprocessminingsystem

Experimental Verification on Deep Learning-based Predictive Process Mining System

Abstract

Over the recent two decades, the research field of process mining has brought great benefits in the development and operation of businesses, especially for business organizations using workflow management systems or information system tools to operate their business activities. Process mining algorithms obtain information and knowledge about what happened in the running phases of businesses through system event logs. One of the important branches of study in process mining is predicting events that will happen when operating a business process, also known as predictive process mining research. Extracting information from the log derives the relationship between events, activities, and resources involved in performing all processes in the organization. This paper introduces a complete predictive process mining system based on deep learning algorithms. More precisely, we introduced the architecture, algorithms and how to install and evaluate this system on some event logs data sets provided by the 4TU research data center. The system promises to be a helpful tool and reference method for researchers interested in the application of artificial intelligence as well as deep learning in creating predictive process mining systems.

Folder structure:

• Main folders: Main project
• How to train and prediction: Using 'TrainandPrediction/Main' folder

How to install the system:

• Install requirement packages
• Run command: Python manage.py run server

Images:

• System architecture:
• System UI:

• Trained models: See 'TrainandPrediction/TrainedModel'
• Prediction results: