Predicting indoor spatial movement using data mining and movement patterns

The ability to accurately predict the movement trajectory of people holds potential benefits for many applications, such as aged care and retail. Such movement predictions rely on collecting and analyzing large amounts of positioning data from sensors. In this work, we describe new algorithms to mine and predict people's movement in an indoor environment. Movement patterns are mined from historical positioning data, and the patterns are used to construct a probability tree which is a visual representation of frequent movements. We have conducted an empirical study in a staff tearoom to capture positioning data, mine movement patterns and construct a probability tree. We show the predictive power of the algorithms using different trajectory estimation strategies.