mlsdm: Machine Learning based Species Distribution Modeling Library for Python > 3.6

mlsdm is a Python library for species distribution modeling (SDM) using machine learning and geospatial data processing. It provides tools for handling raster data, generating pseudo-absence points, training machine learning models, and visualizing results, making it ideal for ecological niche research and modeling. More information can be found in the Vignette document. An example use case with code and context can be found in the example notebook.

Capabilities

• Raster Processing: Clip rasters to presence extents, stack rasters, and convert to DataFrames.

• Pseudo-Absence Sampling: Generate pseudo-absence points using PCA+KDE or random sampling methods.

• Machine Learning: Train and evaluate ensemble SDModels (RandomForest, ExtraTrees, XGBoost, LightGBM).

• Hypertuning: Hypertune each individual model

• Visualization: Plot rasters, partial dependence plots, and presence vs. pseudo-absence points.

• Feature Analysis: Analyze feature importance and perform recursive feature elimination (RFE) with Cross-validation (CV).

mlsdm/
├── mlsdm/
│   ├── __init__.py
│   └── mlsdm.py
├── tests/
│   ├── __init__.py
│   └── test_mylibrary.py
├── example/
│   ├── example.ipynb
│   └── example_data/
|       |── raw_current_rasters/
|       |── raw_future_rasters/
|       |── presence/
|       |── outputs/     
├── README.md
├── Vignette.md
├── setup.py
├── LICENSE
└── requirements.txt

Installation

pip install -r requirements.txt
pip install mlsdm

OR

git clone https://github.com/yourusername/sdmtools.git
cd sdmtools
pip install .

Dependencies

requirements.txt

pip install -r requirements.txt
sudo apt-get install libgdal-dev #in case of error

Testing

python -m unittest discover tests

Contact

For questions or issues, please open an issue on the GitHub repository. For other enquires contact EpiPandit Lab at UC Davis