Activity Behavior Pattern Mining and Recognition

Abstract

As human activity in mobile environments is facing with an ever-increasing range of data, therefore, a deeper understanding of the human activity behavior pattern and recognition is of important research significance. However, human activity behavior that consists of a series of complex spatiotemporal processes is hard to model. In this paper, we develop a platform to do pattern mining and recognition, the main work is as follows: (1) For comparing activity behavior, similarity matrix is computed based on activity intersection, temporal connections, spatial intersection, participant intersection and activity sequence comparison. (2) For calculating activity sequence similarity, an algorithm with O(p(m – p)) is proposed by line segment tree, greedy algorithm and dynamic programming. (3) Activity behavior pattern and socio-demographic pattern are derived by clustering analysis and mining. (4) Pattern is recognized under the inter-dependency relationship between activity behavior pattern and socio-demographic pattern.