Outlier Detection Method of Three Rate Value Based on One Class SVM

Abstract

The three rate value of raw meal is one of the important factors affecting cement quality in production process. The fluctuation range of the three rate value of raw meal has a great influence on the quality of cement clinker. How to identify the outliers of the three rate value has become a key research issue. In this paper, the outlier detection method of three rate value based on One Class SVM (OCSVM) was studied. Firstly, the importance of abnormal value detection of three rate value was introduced. Then, the principle of outlier detection based on OCSVM was described. On this basis, the outlier detection model of three rate value based on OCSVM was introduced. Finally, the proposed method was tested in a cement plant. The test results showed that OCSVM can detect the outliers of the three rate values. This paper provides a method to detect the outliers of the three rate values.