Code and data for the ClusPath algorithm.
Matlab M Objective-C
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
code
data
example output
paper
.octaverc
LICENSE
README.md

README.md

Purpose:

This is the distribution of the ClusPath code. It is distributed under the GNU General Public License v3.0. This code was developed as a research prototype and it should be treated as such. It comes with no guarantees or support, however questions can be addressed to Marian-Andrei@rizoiu.eu.

Contains:

The ClusPath source code and two datasets (CPDS1 dataset and European Companies). For the description of the ClusPath algorithm, please check the paper:

Marian-Andrei Rizoiu, Julien Velcin, Stéphane Bonnevay, & Stéphane Lallich. ClusPath: A Temporal-driven Clustering to Infer Typical Evolution Paths. Data Mining and Knowledge Discovery (DAMI), 30(5), pp. 1324-1349, (2016). paper here

Requirements:

The current code was developed and tested under Linux using Octave v3.8.1, with the following packages installed:

  • control
  • econometrics
  • general
  • io
  • missing-functions
  • nan
  • optim
  • struct
  • statistics

Install additional tools for plotting (in Ubuntu): $> sudo apt-get install graphviz epstool

Octave is a fast-changing environment and new functionalities appear constantly, while old functionalities might be retired without notice. Workarounds might be required to make ClusPath function with the newest version of Octave.

Example usage:

Start octave in the main folder of the project. Upon lauching octave, the ClusPath code is loaded into the Octave path (check the .octaverc file).

$> octave

Next, execute ClusPath with parameters: Alpha = 0.95, Beta = 0.0002, Delta = 3, lamb1 = 1, lamb2 = 10, lamb3 = 1000

octave:2> [Adj, clusassign, dtimestamp, centroids, ctimestamp] = temporalKMeans('file', 'data/cpds1.csv', 'typeT', 5, 'Alpha', 0.95, 'Beta', 0.0002, 'Delta', 3, 'lamb1', 1, 'lamb2', 10, 'lamb3', 1000, 'saveResults', true );

Evaluate the clustering using the four measures:

octave:3> evaluateTemporalClustering('inputfile', 'data/cpds1.csv', 'clusassign', clusassign, 'centroids', centroids, 'ctimestamp', ctimestamp, 'Alpha', 0.95);