Predicting activity occurrence time in smart homes with evolving fuzzy models

In this paper, we address the problem of predicting the time of occurrence of next activity, given the current activity and the context. The models that predict activity and time of occurrence rely on the basic idea that human beings perform sequence of activities at specific times regularly. In other words, the models are dependent on human behavior. However, human behavior changes over time. Also, due to demands and goals to be attained, there may be change in human behavior. Therefore, one of the essential requirements of the predictive models for the given task is autonomous adaptation with time and without undergoing any retraining. Considering the requirement of an adaptive model, we propose an evolving fuzzy rule-based predictive model that can autonomously adapt with changes in the human behavior. The performance of the model is evaluated using real-life data with evolving characteristics and satisfactory results are obtained.