Code and Model Description for Li and Ga Prediction in Coal Gangue
Description
This submission includes the source code and results for the comparative study of machine learning models aimed at predicting the concentrations of lithium (Li) and gallium (Ga) in coal gangue based on major elemental compositions. The models evaluated include: K-Nearest Neighbors (KNN) Support Vector Regression (SVR) Decision Tree Bagging Random Forest Extra Trees Gradient Boosting XGBoost Key Visualizations: Experiment vs. Prediction Comparison: For each model, the actual versus predicted values of Li and Ga content are compared using scatter plots, demonstrating model performance on both training and testing datasets. Partial Dependence Plot (PDP): Generated for the best-performing model to show the influence of major elemental features on the predictions of Li and Ga content. SHAP Summary and Force Plots: SHAP (SHapley Additive exPlanations) plots are provided to illustrate feature importance and explain individual predictions made by the models, further detailing the contribution of each major element The data used are located in the supporting files.