Gold cyanide leaching recovery prediction model based on neighbourhood component analysis and artificial intelligence technique

The complex interaction between key process parameters and their influence on gold cyanide leaching recovery is a challenging problem for mining companies. This paper proposes a gold cyanide leaching recovery prediction model based on the combination of neighbourhood component analysis (NCA) and the artificial intelligence (AI) method. NCA was employed as a dimensionality reduction technique to select relevant input features that contribute significantly to gold cyanide leaching recovery. Thus, the NCA-selected features served as input variables in the applied AI methods, including the backpropagation neural network (BPNN), radial basis function neural network (RBFNN), and light gradient boosting method (LGBM). The input variables were used to develop hybrid prediction models of NCA-BPNN, NCA-RBFNN, and NCA-LGBM. For robust performance evaluation, the proposed hybrid models were compared with variant models combining principal component analysis (PCA) with BPNN, RBFNN, and LGBM to form PCA-BPNN, PCA-RBFNN, and PCA-LGBM. The study further compared the proposed hybrid NCA and AI-based models with the standalone methods of BPNN, RBFNN, and LGBM. The results showed that the prediction performance of the standalone AI methods is improved when combined with the NCA. Inter-comparison between the models revealed that the NCA-BPNN model is the most robust and efficient computational tool for predicting gold cyanide leaching recovery. The reason is that the NCA-BPNN achieved the best performance metrics (RMSE, R, MAPE and R²) and the least Akaike information criterion value in training, validation, and testing stages.