Title: SEQUENCE ALIGNMENT BASED PROCESS FAMILY EXTRACTION
PhD Candidate: Eren Esgin
Program: Informations Systems Department
Date: 17 December 10:00
Place: Conference Hall-01
Abstract: Business Process Management (BPM) paradigm gains growing attention by generic process design and execution capabilities empowered by process-aware information systems. During execution of these transactional information systems, end-users leave traces in the form of event logs, which can be used as a main data source for end user behavior analysis. Process mining encompasses the techniques for automated process discovery from these event logs, conformance checking between the reference process model and process executions, as well as analyzing, predicting and enhancing the performance of business processes. With the emergence of new shared economical models and system architectures, monolithic process perspective is evolved through cross-organizational applications. While contemporary information systems provide functionality for process management within the organizations, a systematic approach to support and analyze multiorganizational processes is missing. Cross-organizational process mining supports the use of commonality and collaboration for process configuration. However, this functionality surplus creates the challenge of dealing with variability across organizations. In this study, we propose a three phased cross-organizational process mining framework in order to extract the commonalities among different organizations serving the same business values. While dominant behavior extraction phase initially derives the sequence of tasks expressing the most typical behavior within the process instances, sequence alignment phase measures the degree of similarities between the process candidates by confidenceenhanced cost functioning, and depicts the neighborhood among these alternatives in terms of process family tree. At process configuration phase, common regions that indicate a functional inheritance or abstractions in the process families are visualized at sequence alignment matrices and interpreted by new feature sets, namely identical and maximal identical pair. According to the experimental results, proposed approach presents a viable and robust cost function in incorporating the business context at process similarity measurement and clustering the process alternatives into process families.