Outlier Mining in Real-time Measurement Data of Sensor Based on Data Mining Technique

摘要

In this paper, the reasons and necessities of the outlier detection in the real-time measurement data of sensor was discussed. Comparing with the main current algorithmic of outlier mining, making use of inherent relations of time sequence in real-time measurement data, using the process mechanism models of outlier detection, using technique of state estimate, the online methods of scooping out the outlier of sensor measurement data are presented. The principle of the outlier detection is computing reliability and the threshold of the real-time data by sensor measurement. Simulation researches in computer outlier miming of measurement of sensor data reveal that this method of outlier detection is valid.