A Unified Framework for Modeling and Predicting Going-Out Behavior

Living in society, to go out is almost inevitable for healthy life. There is increasing attention to it in many fields, including pervasive computing, medical science, etc. There are various factors affecting the daily going-out behavior such as the day of the week, the condition of one’s health, and weather. We assume that a person has one’s own rhythm or patterns of going out as a result of the factors. In this paper, we propose a non-parametric clustering method to extract one’s rhythm of the daily going-out behavior and a prediction method of one’s future presence using the extracted models. We collect time histories of going out/coming home (6 subjects, total 827 days). Experimental results show that our method copes with the complexity of patterns and flexibly adapts to unknown observation.