Mining Moving Patterns Based on Frequent Patterns Growth in Sensor Networks

A novel algorithm named FP-mine (FP: Frequent Pattern) is proposed in this paper to mine frequent moving patterns with two dimensional attributes including locations and time in sensor networks. FP- mine is based on a novel data structure named P-tree and an algorithm of frequent pattern growth named FP-growth. The P-tree can efficiently store large numbers of original moving patterns compactly. The algorithm FP-growth adopts an idea of pattern growth and a method of conditional search, recursively fetches frequent prefix patterns from the conditional pattern bases directly, and joins the suffix to make a pattern grow. Simulation results show FP-mine can efficiently discover frequent moving patterns with two dimensional attributes in sensor networks and decreases its time and space complexity simultaneously.