st-clustering

Package containing various spatio-temporal clustering algorithms


License
MIT
Install
pip install st-clustering==0.0.5

Documentation

ST-CLUSTERING

st_clustering is an open-source software package for spatial-temporal clustering:

  • Built on top of sklearn's clustering algorithms
  • Scales to memory using chuncking. See the st_fit_frame_split method

Installation

The easiest way to install st_clustering is by using pip :

pip install st_clustering

How to use

import st_clustering as stc

st_dbscan = stc.ST_DBSCAN(eps1 = 0.05, eps2 = 10, min_samples = 5)
st_dbscan.st_fit(data)
  • Demo Notebook: this Jupyter Notebook shows a demo of common features in this package.

Description

A package that implements a straightforward extension for various clustering algorithms to accomodate spatio-temporal data. Available algorithms are:

  • ST DBSCAN
  • ST Agglomerative
  • ST OPTICS
  • ST Spectral Clustering
  • ST Affinity Propagation
  • ST HDBSCAN

For more details please see original paper:

Cakmak, E., Plank, M., Calovi, D. S., Jordan, A., & Keim, D. (2021). Spatio-temporal clustering benchmark for collective animal behavior. In 1st ACM SIGSPATIAL International Workshop on Animal Movement Ecology and Human Mobility (HANIMOB’21).

License

Released under MIT License. See the LICENSE file for details.