Situation-Aware Decision Making in Smart Homes

The ability to efficiently predict the elderly's future situations and make the right decision accordingly is a necessity in developing smart homes. In this paper, we propose a hybrid and dynamic predictive model which utilizes higher order Markov models integrated with a situation ranking technique. More specifically, we employ a revised version of PageRank algorithm to take the properties of the situation-graph (structure and semantics) into account and dynamically rank the situations considering the user's mental state. Then we apply rankings as prior probabilities in order to build the corresponding Markov model. Also, we utilize rankings to identify milestone situations and transitions in order to compress the representation model. Experiments show that the predictions obtained by this approach are more efficient and effective than the ones produced from the pure predictive graphical model-based approaches.