Code and Model Description for Li and Ga Prediction in Coal Gangue

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.

本提交包含针对基于主要元素组成预测煤炭矸石中锂（Li）和镓（Ga）浓度的机器学习模型的比较研究之源代码与结果。所评估的模型包括：

K-最近邻算法（KNN）、
支持向量回归（SVR）、
决策树、
Bagging、
随机森林、
Extra Trees、
梯度提升、
XGBoost。

关键可视化包括：

实验与预测对比：通过散点图比较每个模型的实际与预测的锂和镓含量值，展示了模型在训练集和测试集上的性能。

局部依赖性图（PDP）：为表现最佳的模型生成，以展示主要元素特征对锂和镓含量预测的影响。

SHAP 概述与影响图：提供 SHAP（SHapley Additive exPlanations）图来展示特征重要性并解释模型做出的单个预测，进一步详细说明每个主要元素的贡献。

所使用的数据位于附加文件中。