Skip to content

README.md

Latest commit

 

History

History
164 lines (119 loc) · 4.19 KB

README.md

File metadata and controls

164 lines (119 loc) · 4.19 KB

Geochem Classifier 🧪⛏️

A machine learning–powered geochemical classification tool designed to analyze and predict the type of mining samples.

📋 Overview

This project leverages machine learning algorithms to automatically classify geochemical samples into three categories:

  • Sterile – No economic interest
  • Potential – Moderate mining potential
  • Ore – High economic value

The system bases its predictions on 36 geochemical features.

🚀 Key Features

  • Automatic classification of geochemical samples
  • Threshold analysis to refine classification criteria
  • Smart preprocessing with handling of missing or censored values
  • Algorithm benchmarking (SVM, Random Forest, etc.)
  • Real-time predictions on new samples
  • Detailed reports with performance metrics

📁 Project Structure

src/
├── train.py                  # Model training
├── infer.py                  # Predictions on new samples
├── preprocess.py             # Data preprocessing
├── analyze_ag.py             # Silver-focused analysis
├── analyze_thresholds.py     # Threshold optimization
├── geochemical_analysis.py   # General geochemical analyses
├── potential_analysis.py     # Analysis of potential samples
└── silver_analysis.py        # Silver-specific analysis

🛠️ Installation

Requirements

  • Python 3.8+
  • pip

Install dependencies

pip install -r requirements.txt

Core dependencies

  • scikit-learn – Machine learning algorithms
  • pandas – Data handling
  • numpy – Numerical computations
  • joblib – Model serialization
  • matplotlib – Data visualization

💻 Usage

1. Train a model

python src/train.py

This trains multiple models and saves the best one under models/geochem_pipeline.joblib.

2. Run a prediction

from src.infer import predict_sample

# Example input (36 geochemical values)
sample = [57.86, 16.46, 7.96, 0.92, 3.39, 3.29, 0.11, 0.74, 0.23, 2.77,
          5.5, 1400, 117, 1287, 0.9, 20, 4, 91, 119, 123, 10, 3, 49, 12,
          43, 68, 65, 1.76, 9, 40, 20, 110, 23, 17, 164, 4.46]

prediction = predict_sample(sample)
print(f"Classification: {prediction}")

3. Run threshold analysis

python src/analyze_thresholds.py

4. Specialized analyses

# General analysis
python src/geochemical_analysis.py

# Silver analysis
python src/silver_analysis.py

# Potential sample analysis
python src/potential_analysis.py

📊 Features Analyzed

The model uses 36 geochemical parameters, including:

Major elements (%)

SiO₂, Al₂O₃, Fe₂O₃, MgO, CaO, Na₂O, K₂O, TiO₂, P₂O₅, MnO

Trace elements (ppm)

Cu, Pb, Zn, Ag, Au, As, Sb, Bi, Cd, Co, Cr, Ni, Mo, W, Sn, V, Ba, Sr, etc.

🎯 Model Performance

Typical results:

  • Accuracy: >85%
  • Recall: >80%
  • F1-score: >82%

Detailed metrics are stored in the training reports.

📋 Data Format

Samples must be provided as a list of 36 values in the exact expected order.
The preprocessing step automatically:

  • Converts decimal commas → points
  • Handles values like <0.1 → converted to half (0.05)
  • Fills missing values → 0.0

🔧 Configuration

Configurable parameters in the scripts:

  • use_raw_data=True → Enables preprocessing of raw input
  • test_size=0.3 → Share of data reserved for testing
  • random_state=42 → Seed for reproducibility

📈 Example Output

TEST - Ore Sample:
Normalized data:
Min: -1.234
Max: 2.156
Mean: 0.045

Class probabilities:
  Sterile:   0.123
  Potential: 0.234
  Ore:       0.643

Result: Ore

🤝 Contributing

Contributions are welcome!

  1. Fork the repository
  2. Create a feature branch (git checkout -b feature/new-feature)
  3. Commit your changes (git commit -m 'Add new feature')
  4. Push your branch (git push origin feature/new-feature)
  5. Open a Pull Request

📝 License

This project is under the MIT License. See the LICENSE file for details.

📞 Contact

For questions or suggestions, please open an issue on GitHub.

⭐ If you find this project useful, don’t forget to give it a star!