A Grounding Grid Model of Corrosion Factors Based on Clustering and SVR

Abstract

A structure combined clustering analysis and Support Vector Regression (SVR) was put forward with to establish a grounding grid model of corrosion factors. Firstly, according to the determination of clustering numbers required by K-average algorithm, a new method to determine clustering numbers was proposed with the distance between intra-class and inter-class as the clustering effectiveness measuring; secondly, a training sample set after cluster analysis based upon Kopt-average algorithm was divided into class clusters with similar characteristics; finally, for each cluster after clustering, corresponding models were established respectively. The result shows that the K- average algorithm can determine the best clustering number of a sample set effectively, and the grounding grid SVR model of corrosion factors is not only effective, but also steadily increases the corrosion rate prediction accuracy.