An Outlier Detection Method and Its Application for Earthen Ruins Data Monitored by WSN

This paper proposed an outlier detection method to detect events that caused by environmental anomalies in earthen ruins monitoring data using wireless sensor networks (WSN). The proposed method, Outlier Detection method based on a SVR-based regression prediction method combined with a distance-based Voting strategy (ODSV), first uses support vector regression (SVR) to predict the earthen ruins monitoring data, and then a threshold based on data characteristics is used to distinguish whether the data is a normal one or an outlier. Voting strategies based on distance weights are used to distinguish between errors and events in the outliers. The experiments have been designed and implemented on the monitoring data of Tang Dynasty Daming Palace and Ming Dynasty Great Wall. The experimental results show the good outlier detection accuracy of the proposed method.