Data and Code Documentation

Data S1 (separate file)

(GaussianData.xlsx) contains a total of 4,747 data points that adhere to a Gaussian distribution. This file includes: 2 columns of location coordinates, 4 columns of features, 1 column of true labels, and training labels across 5 different experimental conditions.

Data S2 (separate files)

(BYEB.xlsx) contains the raw data. (Processed_BYEB.xlsx) is the preprocessed version, which also includes 4,747 data points. This file comprises: 2 columns of location coordinates, 4 columns of physical property features (resistivity, magnetization, density, and velocity), 1 column of labels, and labels under 4 different experimental conditions.

Code

This MATLAB script performs lithology classification, executes dimensionality reduction, trains a dual-constrained semi-supervised machine learning method (DCSM), and evaluates the model’s performance. Additionally, it visualizes the results of model training and convergence.

1. GaussianData (GaussianData.xlsx)

This MATLAB script implements the DCSM to identify and classify the multivariate Gaussian simulation data. There are notable differences in the evaluation metrics between Experiments 1-4 and Experiment 5. We provide two demos to illustrate the results.

• Run demo1_Gaussian (Experiment 5): This demo showcases the results obtained when training the model using 6 pseudo-boreholes.
• Run demo2_Gaussian (Experiments 1-4): This demo retrieves the results for Experiments 1-4. Please modify the following lines accordingly:

  Experiment 1: training_label = data.LABLE2_4';
  Experiment 2: training_label = data.LABLE3_3';
  Experiment 3: training_label = data.LABLE4_2';
  Experiment 4: training_label = data.LABLE5_1';
                  

2. BYEB (Processed_BYEB.xlsx)

• Run demo_BYEB (Experiments 1-4): This MATLAB script utilizes the DCSM to identify and classify deep-seated mineralized dolomite in the Bayan Obo mining area, Inner Mongolia, China. This demo retrieves the results for Experiments 1-4. Please modify the following lines accordingly:

  Experiment 1: training_label = data.LABLE2_4';
  Experiment 2: training_label = data.LABLE3_3';
  Experiment 3: training_label = data.LABLE4_2';
  Experiment 4: training_label = data.LABLE5_1';