Local outlier detection method towards data stream

In view of the present outlier detection method is only suitable for static environment, but couldn´t timely response to the dynamic characteristics of data streams. This paper puts forward a dynamic adaptive test method over outlier of data stream according to the characteristics of data streams. This method based on density of local anomalies detection algorithm LOF, Uses multi-window sliding technical to judge data stream of outlier through analyze local anomalies of data stream, through improving calculation method of local anomalies factor, so, users can set the initial threshold to control the overall detection rate and the rate of false positives. Experiment results show that this method is effective.