This paper considers the problem of outlier detection in data stream, proposes a new metric called weighted frequent pattern outlier factor for categorical data streams, and presents a novel fast outlier detection algorithm named FODFP-Stream (fast outlier detection for high dimensional categorical data streams based on frequent pattern). FODFP-Stream computes the outlier measure through discovering and maintaining the frequent patterns dynamically, and can deal with the high dimensional categorical data streams effectively. FODFP-Stream can also be extended to resolve continuous attributes and mixed attributes data streams. The experimental results on synthetic and real data sets show the promising availabilities of the approaches.