2018 : Anomaly Detection Based on Control-flow Pattern of Parallel Business Processes

Prof. Drs. Ec. Ir. Riyanarto Sarno M.Sc Ph.D


Abstract

Company must have an audit trail that captures activities executed, named an event log. In executing the process, there are certain gap between what is expected and what is executed, called an anomaly. Anomaly have to be evaluated so it does not harm the company. Process mining is implemented to model actual workflow. Anomaly contained in the event log caused the low fitness and precision value. In this research, trace clustering is implemented to group the same trace into one cluster. Then, data filtering is done to filter traces which have a low frequency value. After that, the event log is modeled to form a new business process. The filtered business process model has higher fitness and precison value compared to unfiltered business process. The fitness and precision of unfiltered process model is 1 and 0.6446991. Meanwhile, the fitness and precision of filtered process model is 1 and 0.81481093.