A Conceptual Framework for Supporting Deep Exploration of Business Process Behavior

Abstract

Process mining serves for gaining insights into business process behavior based on event logs. These techniques are typically limited to addressing data included in the log. Recent studies suggest extracting data-rich event logs from databases or transaction logs. However, these event logs are at a very fine granularity level, substituting business-level activities by low-level database operations, and challenging data-aware process mining. To address this gap, we propose an approach that enables a broad and deep exploration of process behavior, using a conceptual framework based on three sources: the event log that holds information regarding the business-level activities, the (relational) database that stores the current values of data elements, and the transaction (redo) log that captures historical data operations performed on the database as a result of business process activities. Nine types of operations define how to map subsets of elements among the three sources in order to support human analysts in exploring and understanding the reasons of observed process behavior. A preliminary evaluation analyzes the outcomes for four useful scenarios.