Improving the Performance of Action Prediction through Identification of Abstract Tasks

摘要

An intelligent home is likely in the near future. An important ingredient in an intelligent environment such as a home is prediction - of the next action, the next location, and the next task that an inhabitant is likely to perform. In this paper we describe our approach to solving the problem of predicting inhabitant behavior in a smart home. We model the inhabitant actions as states in a simple Markov model, then improve the model by supplying it with data from discovered high-level inhabitant tasks. For simulated data we achieved good accuracy, whereas on real data we had marginal performance. We also investigate clustering of actions and subsequently predict the next action and the task with hidden Markov models created using the clusters.