An intelligent approach to data extraction and task identification for process mining

Business process mining has received increasing attention in recent years due to its ability to provide process insights by analyzing event logs generated by various enterprise information systems. A key challenge in business process mining projects is extracting process related data from massive event log databases, which requires rich domain knowledge and advanced database skills and could be very labor-intensive and overwhelming. In this paper, we propose an intelligent approach to data extraction and task identification by leveraging relevant process documents. In particular, we analyze those process documents using text mining techniques and use the results to identify the most relevant database tables for process mining. The novelty of our approach is to formalize data extraction and task identification as a problem of extracting attributes as process components, and relations among process components, using sequence kernel techniques. Our approach can reduce the effort and increase the accuracy of data extraction and task identification for process mining. A business expense imbursement case is used to illustrate our approach.