Prediction of silicon content in hot metal based on SVM and mutual information for feature selection

Yikang Wang,Xiangguan Liu

2011-01-01

Journal of Information and Computational Science

Abstract:A prediction model was proposed based on support vector machine (SVM) and mutual information (MI) for feature selection. MI criterion is applied to evaluate a set of candidate features and to select an informative subset to be used as input variables for SVM. Based on MI, the proper input variables, which describe the multivariate time series dynamics properly, will be selected. The selected inputs have a maximum relevance to output variable and there exists minimum redundancy between them. Presented results show that the SVM model based on MI for feature selection has better performance compared with SVM without feature selection. The hit rate reaches 83% in test set, much better than the commonly used BP working with the same set. The proposed approach seems promising and can be determinant in providing the experts with the right tools for the selection of the relevant factors and for the prediction in this difficult problem, and it can satisfy the requirements of on-line prediction of silicon content in hot metal. © 2009 by Binary Information Press.