Abstract

Mineral resource prediction is becoming increasingly important as researchers attempt to resolve the prospect direction by mining geological data. In this paper, Support Vector Regression (SVR) is applied to predict iron deposit metallogenic favourability degree since SVR is a powerful tool to solve the problem characterized by smaller sample, nonlinearity, and high dimension with a good generalization performance based on structural risk minimization. The paper discusses the support vector regression algorithm in some detail, describes a SVR based-system that learns from examples to predict metallogenic favourability degree of iron deposit and contrasts this approach with Partial Least Squares (PLS). The experimental results show that SVR has high recognition rates and good generalization performance for small sample, especially good for treating the data of some nonlinearity in geology.

Key words: Support Vector Regression (SVR), metallogenic favourability degree, mineral resource, quantitative prediction.