Treatment of lack of information in process mining

Abstract

Process mining is a research discipline including discovery, monitoring and improvement of real processes by extracting knowledge from event logs readily available in today's information systems. Most process mining algorithms assume that the traces are complete and free of noise. In reality, this assumption is rarely met. The lack of information affects the structure and understanding of the model discovered. The paper proposes a set of steps to estimate the lack of information in the traces used in processes mining. Align traces during the pre-processing stage is proposed as part of the estimation in order to detect possible situations in which lack of information is manifested. From detected situations, missing information is estimated and a new record is generated in correspondence to the event, which can be used by different process discovery algorithms. Finally we discuss the experimental results obtained from implementing the approach.