Efficient Mining of Outlying Sequential Behavior Patterns

Yifan Xu, Lei Duan, Guicai Xie, Min Fu, Longhai Li, Jyrki Nummenmaa

Abstract: Sequential patterns play an important role when observing behavior. For instance, the daily routines and practices of people can be characterized by sequences of activities. These activity sequences, in turn, can be used to find exceptional and changed behavior. Observing students’ behavior changes is an effective approach to find indications of mental health problems, and changes in an elderly person’s daily activities may indicate a weakening health condition. With the availability of behaviour sequential events, outlierness analysis of behavior sequences has been established as a meaningful research problem. This paper considers the mining of outlying behavior patterns (OBP) from sequential behaviors. After discussing the challenges of OBP mining, we present OBP-Miner, a heuristic method that computes OBPs by incorporating various pruning techniques. Empirical studies on two real-world datasets demonstrate that OBP-Miner is effective and efficient.