Melhoria de modelos de processo de negócio com mineração de processos e simulação baseada em agentes

Abstract

Process mining techniques allow the discovery of process models in event logs. Despite the signicant progress in this eld, the mined process models are often presented in a complex form, diculting the understand of business processes. This complexity involves implicitly the relationship between the high-level abstraction used by analysts to describe their business processes, and the low-level activities, represented in the event logs. In fact, there is a gap between the low-level events and the high-level abstraction of business activities. This research addresses this problem by developing a process mining technique so called hierarchical Markov model. This model is able to produce a hierarchical process model relating the low- and high-level events through the use of a discovering procedure based on the Expectation-Maximization technique. For the event logs generation closest to the organizational reality, where agents are involved in dierent business scenarios, we used the Agent-Object Relationship (AOR) simulation framework. In order to evaluate the proposed approach three dierent business process models were described using the Business Process Modeling Notation, together with the agent-based simulation platform (AOR) to generate the event logs, and also implemented a case study using a real-life event log. The results were compared with existing process mining techniques in the ProM framework. Also, metrics were studied and selected to quantitatively evaluate aspects of complexity in the generated hierarchical process models. The evaluation results indicated the feasibility of the hierarchical proposed model to ll the gap between the two business process abstraction levels existents in organizations.