PRAr

Partition Relevance Analysis with the reduction step

This repository contains supporting material for the paper Weighted Cluster Ensemble Based on Partition Relevance Analysis With Reduction Step (work in progress).

Requirements

• MATLAB
• Python3

Code was tested with MATLAB R2019b and Python 3.8.2 on Windows 10. Please, give us feedback if you experience any troubles on other configurations.

Installation

• Download and unzip into a folder on your computer.
• Open the folder in MATLAB.
• Run setup.m.

After successful setup, consider the following:

• example/demo_PRAr.m: a demo script showing the main functionalities of PRAr in the context of weighted cluster ensemble;
• experiment/runExperiment.m: script for the reproduction of the results published in the paper.

Pepelka toolbox

The Pepelka toolbox is required to run an example on PRAr and full experiment. Pepelka (means Cinderella in the Slovene language) is a MATLAB toolbox for data clustering and visualization. It provides functions for:

• data loading and preprocessing,
• finding clusters using single-clustering and ensemble methods,
• cluster internal and external validation,
• visualization of clustering results.

Pepelka includes a lot of artificial and real-world datasets.

A pre-release of Pepelka is included here.