Recognize contextual situation in pervasive environments using process mining techniques

Research in pervasive computing and ambience intelligence aims to enable users to interact with the environment in a context-aware way. To achieve this, a complex set of features describing different aspects of the environment has to be captured and processed; in other words situation-awareness is needed. This article notes uniquely three points when modelling situations. Firstly, unlike most existing approaches, context information history should be considered when modelling the situations. We argue here that the current state cannot be understood in isolation from the previous states. Secondly, in order to track user’s behaviour there is a need to consider the context information available in the different domains the user visits. Thirdly, to identify situations it can be problematic to define situation patterns and looking for an exact match as most of the approaches does. We found that the combination of the flexibility of the user behaviour and automated capture of context events provide a very effective solution for contextual situation recognition. In this article we first provide a formalization of the situation recognition problem and then we focus on the potential use of process mining techniques for measuring situation alignment, i.e., comparing the real situations of users with the expected situations. To this end, we propose two ways to create and/or maintain the fit between them: linear temporal logic (LTL) analysis and conformance testing. We evaluate the effectiveness of the framework using a third party published smart home dataset. Our experiments prove the effectiveness of applying the proposed approach to recognizing situations in the flow of context informatio