Confidence-Aware Sequence Alignment for Process Diagnostics

Traditional process modeling in contemporary information systems concentrates on the design and configuration phases, while less attention is dedicated to the enactment phase. Instead of starting with a process design, process mining attempts to discover interesting patterns from a set of real time execution namely event logs, which can be handled as a main data source for end-user behavior analysis, and translate this discovered knowledge into process model. One of the challenging issues in process mining is process diagnostics, i.e. encompassing process performance analysis, anomaly detection, diagnosis, inspection of interesting patterns, and sequence alignment is applicable to find out common subsequences of activities in event logs that are found to recur within a process instance or across the process instances emphasizing some domain significance. In this study, we focus on a hybrid quantitative approach for performing process diagnostics, i.e. comparing the similarity among process models based on the established dominant behavior concept, confidence metric and Needleman-Wunsch algorithm with dynamic pay-off matrix.