Application of RS Theory and SVM in the Ore-Rock Classification

Abstract

One of the main goals in machine learning is the general functional dependencies. Recent advances in kernel- based methods are focused on designing flexible and powerful input and output representations. This paper describes how rough set (RS) and support vector machine (SVM) can be practically implemented in ore-rock classification, and discusses the kernel mapping technique which is used to construct SVM solutions. In ore-rock classification using RS theory and SVM, original sample data is preprocessed with the knowledge reduction algorithm of RS theory, and the redundant condition attributes and conflicting samples are eliminated from the training sample sets to reduce space dimension of the data. Preprocessed data is used as training data of SVM, and fuzzy discrete model is used as training model. The results show that the RS and SVM can improve the training speed and precision of ore-rock classification.