Discovering workflow patterns from timed logs

Abstract

Traditionally, workflow modeling is typically done by interviews or questionnaires which stay quite limited for the acquisition of workflow models and their adaptation to changing requirements. In order to build such model, this paper proposes to “reverse the process”. The approach, we propose, is able to acquire workflow model from workflow log, which contains information about execution events. This paper presents a mining technique to discover workflows patterns (e.g. Sequence, Parallel split, Exclusive choice, etc.) from workflow event log in the objective to build a global workflow model. An event log is a view of a workflow execution at a given instant. The mining of workflow patterns is done by the combination of two complementary techniques: an algorithmic technique and a statistical analysis. The building of the global workflow model is done with bottom-up approach. The discovering technique proposed can deal with some ambiguities where the single use of an algorithmic technique method cannot resolve.