SMOTE-XGBoost using Tree Parzen Estimator optimization for copper flotation method classification

By Hai Dong, Dakuo He, Fuli Wang. 2020

49 citations recorded by the source corpus.

Abstract Classification of the flotation method is an important stage in the design of the flotation process. This study faces the problems of small samples and category imbalance through the following steps: (1) The XGBoost was chosen as the multiple classifier, and the geometric mean of the recall rates was used as the evaluation metric. (2) The proposed evaluation set validation greatly reduced the standard deviation (Std) of the evaluation metrics compared with cross-validation. (3) A training set of minority categories oversampled by the synthetic minority oversampling technique (SMOTE) improved the of classification effect of minority categories. (4) The Tree Parzen Estimator (TPE) was used as a hyper-parameter optimization method and realized better performance of the model. The results show that the mean value and Std of GM were 0.867 and 0.014, respectively, and the recall rates of preferential flotation, partial flotation and mixed flotation were 0.849, 0.831 and 0.922, respectively.

Subfields: Accounting, Artificial Intelligence, Computational Theory and Mathematics

View Paper: https://www.semanticscholar.org/paper/3a88f4e1b73e425c0392665c9297eefd8dd431aa