pyproclus

A Python implementation of the PROCLUS algorithm: PROjected CLUStering for high-dimensional data.

Overview

PROCLUS is a subspace clustering algorithm that finds clusters in different subspaces of high-dimensional data. It automatically identifies relevant dimensions for each cluster, making it particularly effective for datasets where different clusters exist in different feature subspaces.

Requirements

• NumPy
• SciPy
• Matplotlib (for running the examples)
• Cython (optional, for building the Adjusted Rand Index evaluator)

Installation

1. Clone the repository:

   git clone https://github.com/Alessi0X/pyproclus-fork.git
   cd pyproclus-fork

2. (Optional) Build the Cython extension for faster evaluation of the Adjusted Rand Index:

   python setup.py build_ext --inplace

   or

   make

Usage

Check the example files (example01.py, example02.py, example03.py) for usage demonstrations.

Basic usage:

import proclus as prc
import arffreader as ar

# Load data
X, supervision = ar.readarff("data/simple.arff")

# Run PROCLUS
medoids, dimensions, assignments = prc.proclus(X, k=3, l=2, seed=902884)

# Evaluate clustering
accuracy = prc.computeBasicAccuracy(assignments, supervision)

Examples

• example01.py - Simple 2D clustering demonstration
• example02.py - High-dimensional data clustering
• example03.py - Subspace stream clustering with custom parameters
• correlation.py - Analysis of correlation between objective function and Adjusted Rand Index

Implementation Notes

The Adjusted Rand Index evaluation measure is implemented in Cython for efficiency. Pre-generated C code is included in the distribution. This component is optional and only required for computing clustering quality metrics.

Reference

Charu C. Aggarwal, Joel L. Wolf, Philip S. Yu, Cecilia Procopiuc, and Jong Soo Park. 1999. Fast algorithms for projected clustering. In Proceedings of the 1999 ACM SIGMOD international conference on Management of data (SIGMOD '99). ACM, New York, NY, USA, 61-72. DOI=10.1145/304182.304188

License

This project is licensed under the GNU General Public License v3.0 - see the LICENSE file for details.

Acknowledgments

This is a fork with Python 3 compatibility updates. Original implementation by Cassio M. M. Pereira.

Original repository: https://github.com/cmmp/pyproclus