An approach to the misleading action solving in plan recognition

Misleading actions have not been considered in previous researches on plan recognition. It leads to poor recognition results for some special domains. This paper introduces new concepts including reliability, correlativity, and correlated action sequence etc. A novel algorithm is proposed to recognize a misleading action by using correlated action sequences. The proposed algorithm is shown to improve the accuracy of plan recognition. Moreover, it could also be applied to intrusion detection and network security problems.