Multi label feature selection for copper flotation process design

The copper flotation process design is a multi label learning. Before multi label classification and regression, label-specific features for each label need to be selected from the original features. This paper puts forward a multi label feature selection (MLFS) based on domain knowledge and label correlation. In the proposed MLFS, the domain knowledge function applies the knowledge of copper flotation in feature selection, and the label correlation function introduces the Kullback Leibler divergence between labels in feature selection. The experimental results demonstrate that the proposed MLFS has superiority over benchmark multi label feature selection algorithms in the copper flotation backbone process design.