학술논문
Proactive business process mining for end-state prediction using trace features
Document Type
Conference
Source
2021 IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computing, Scalable Computing & Communications, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/IOP/SCI) SMARTWORLD-SCALCOM-UIC-ATC-IOP-SCI SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computing, Scalable Computing & Communications, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/IOP/SCI), 2021 IEEE. :647-652 Oct, 2021
Subject
Language
Abstract
Business processes in the complex real-world environment are heterogeneous and challenging to monitor for any possible discrepancies. Businesses substantially rely on the efficiency of these processes to maintain the quality of services for their customers and wish to ensure that an executing business process is progressing in the desired manner. Although process mining techniques provide adequate information about the process execution, it is vital to maintain the quality of business processes through an automated process prediction system that analyses and provides constructive feedback for process improvement. Techniques in the literature can predict the future outcome of a business process, but they lack empirical information about the behaviour of an executing process instance as compared to the optimum process model. In this paper, we have proposed an online process prediction framework using features generated through process mining techniques. We used a heuristic miner algorithm to discover the process model and performed conformance analysis to generate features presenting the contextual behaviour of the process instance. We selected highly contributing features to predict the outcome of the real-world business process using several machine learning algorithms. Our experimental results showed high accuracy, recall, and F-measure. We compared our technique with a similar technique from literature and showed that our solution is more reliable in process outcome prediction.