Hardgrove grindability index prediction using support vector regression

Abstract

Hardgrove grindability index (HGI) measures the grindability of coal and is a qualitative measure of coal. It is referred to in mining, beneficiation and utilization of coal. HGI of coal depends on the coal composition and there is an interest to predict this property from proximate analysis of coal. In this paper, support vector regression (SVR), a potential machine learning technique is used to develop a non-linear relationship between input proximate analyses of coal with output HGI by training the SVR model with limited measured data and to validate it with the rest of the untrained data. SVR is a promising method and suggests that a smaller data set can be used for training the model than what has been studied earlier using artificial neural network (ANN) techniques, so that the model still validates the remaining data.