https://github.com/daniel-furman/pysdms

A Python class for AutoML spatial classification (designed for Species Distribution Modeling applications).
https://github.com/daniel-furman/pysdms

ecological-modeling geospatial-analysis pycaret species-distribution-modeling

Last synced: 6 months ago
JSON representation

A Python class for AutoML spatial classification (designed for Species Distribution Modeling applications).

• Host: GitHub
• URL: https://github.com/daniel-furman/pysdms
• Owner: daniel-furman
• License: mit
• Created: 2021-04-07T04:37:58.000Z (about 4 years ago)
• Default Branch: pypi-deployed
• Last Pushed: 2021-09-05T22:51:57.000Z (over 3 years ago)
• Last Synced: 2024-11-13T00:56:55.096Z (7 months ago)
• Topics: ecological-modeling, geospatial-analysis, pycaret, species-distribution-modeling
• Language: Python
• Homepage: https://pypi.org/project/PySDMs/
• Size: 26 MB
• Stars: 3
• Watchers: 1
• Forks: 1
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
## PySDMs

[![Build Status](https://travis-ci.com/daniel-furman/PySDMs.svg?branch=pypi-updates)](https://travis-ci.com/daniel-furman/PySDMs)

`from PySDMs import PySDMs`

---

## Example 1: "EcoRisk Forecasts - California" for DAT/Artathon 2021






### Descriptive Stats for Climatic Change between 1985 to 2070 at Species Presences:

Bioclimatic Variable  | Coast redwood % Change | Giant sequoia % Change | Joshua tree % Change

-----|-------|-------|-------

Temperature Annual Mean | +22% | +47% | +24%

Temperature Annual Range | +5% | +4% | +2%

Precipitation Driest Month | -1% | -2% | -7%

* SSP 370 [CMIP6](https://www.worldclim.org/data/cmip6/cmip6climate.html) models for the IPCC6 report.

* Bioclimatic Features from [WorldClim2](https://www.worldclim.org/data/worldclim21.html)

* Species presences from GBIF and carefully cleaned




## Example 2: Probablistic near-current interpolation

* Blending methods boosted model performances to ~ two-zero false negatives per species.

**Coast redwood** SDM geo-classification (*Sequoia sempervirens*) | Standard deviations from multiple seeds/samples. 

:---------------------------------:|:----------------------------------------:

![](examples/coast_redwoods/curr-cr.png) | ![](examples/coast_redwoods/current-sd.png)

**Giant sequioa** SDM geo-classification (*Sequoiadendron giganteum*) | Standard deviations from multiple seeds/samples.

:---------------------------------:|:----------------------------------------:

![](examples/giant_sequoias/curr-gs.png) | ![](examples/giant_sequoias/curr-sd.png)

**Joshua tree** SDM geo-classification (*Yucca brevifolia*) | Standard deviations from multiple seeds/samples. 

:---------------------------------:|:----------------------------------------:

![](examples/joshua_trees/curr-jtree.png) | ![](examples/joshua_trees/curr-sd2.png)

## Bio

An object-oriented Python class for semi-auto ML geo-classification (running on PyCaret). Compares gradient boosted tree algorithms by default, with options to include soft voters and NNs. Designed for Species Distribution Modeling applications.

## Package Layout

* [PySDMs](https://github.com/daniel-furman/PySDMs/tree/main/PySDMs)/ - the library code itself

* [LICENSE](https://github.com/daniel-furman/PySDMs/blob/main/LICENSE) - the MIT license, which applies to this package

* README.md - the README file, which you are now reading

* [requirements.txt](https://github.com/daniel-furman/PySDMs/blob/main/requirements.txt) - prerequisites to install this package, used by pip

* [setup.py](https://github.com/daniel-furman/PySDMs/blob/main/setup.py) - installer script

* [tests](https://github.com/daniel-furman/PySDMs/tree/main/test)/ - unit tests

## Functions

   **self.fit():** Model training with PyCaret, considering tree-based

        methods, neural nets, and best-subset-selection soft voting blends.

        Requires a data-frame with a classification target and numerical

        explanatory features. Returns the voter with the best validation

        metric performance (default metric=F1).

   **self.interpolate():** Geo-classification function for model interpolation to

        raster feature surfaces. Saves to file both probabilistic and binary

        distribution predictions.

   **self.validation_performance():** Metric scores and AUC visuals on the test set.