Automated Procedures for Improving the Accuracy of Sensor-Based Monitoring Data

摘要

In this paper, we describe an automated method for detecting and correcting outliers in a data set generated by a weigh-in-motion scale. We use clustering and regression to identify outliers in an aggregated version of the data set. Then we compare the attribute distributions of outliers and the distributions of normal data to identify the mechanisms that cause the outliers. After we remediated these mechanisms, the outlier data closely resembled the normal data.