Workflow Mining Application to Ambient Intelligence Behaviour Modelling
Description of processes that allow developing strategies with quality and in an efficient way is not an easy task. To help the process designers, powerful specication and implementation tools to standardize general processes, and particularly Clinical Guidelines, are needed. The use of Workflows help design experts in the creation of system execution rules in the way computer program- mers write a program. However, the design problem is not solved because of the actual mutability of actual processes, what makes diffcult to design experts know which are the current working processes. The use of Pattern Recognition techniques allows experts learning the control flow inherent to proceses and helps them knowing what is happening with the processes in reality. This methodology is known as Workflow Mining. Clinical Pathways present state changes in their care processes that are triggered by actions results. Existing Workflow Mining methodologies do not allow inferring this information because it is not included in their corpora. Complex problems such as the Clinical Pathways case are not addressed by the current mining systems. In this work, current Workflow Representation, Interpretation and Learning models are analyzed. This study intends to propose an adequate model to solve the problems that prevent Clinical Guidelines designers to access Workflow Mining facilities. For that, a new methodology for an adequate Clinical Pathways design as well as new tools to allows the implantation of this methodology are presented. As a result, this document presents: a new Workflow Mining methodology, a new Workflow representation model with high expressivity and legibility, a new software tool able to execute and simulate Workflows, and a new algorithm able to infer Workflows from past samples. Those results are focused on helping in the design Clinical Pathways.