Combined Fuzzy State Q-learning algorithm to predict context aware user activity under uncertainty in assistive environment

Abstract

In an Assistive Environment (AE), where dependant users are living together, predicting future User Activity is a challenging task and in the same time useful to anticipate critical situation and provide on time assistance. The present paper analyzes prerequisites for user-centred prediction of future Activities and presents an algorithm for autonomous context aware User Activity prediction, based on our proposed combined Fuzzy-State Q-learning algorithm as well as on some established methods for data-based prediction. Our combined algorithm achieves 20% accuracy better than the Q-learning algorithm. Our results based real data evaluation not only confirm the state of the art of the value added of fuzzy state to decrease the negative effect of uncertainty data trained by a probabilistic method but also enable just on time assistance to the User. © 2008 IEEE.