statement.txt

Apply at least three different algorithms of community detection to the attached networks. It is not necessary to implement them, you may use any free access programs you like. At least one of the algorithms must be based on the optimization of modularity, and you must use at least two different programs (i.e. do not use all algorithms from the same program).

Some of the provided networks come with a reference partition, obtained from real information of the network. In those cases, you must compare your partitions with the reference ones, using at least the following standard measures: Jaccard Index, Normalized Mutual Information and Normalized Variation of Information. Once again, you don't need to implement them, they can be calculated e.g. using Radatools.

The delivery must include:

Brief description of the algorithms and programs used.
Selected parameters for each algorithm and/or network, and the used scripts (if any)
The partitions in Pajek format
Plots of the networks with their partitions
A table with the comparison measures
The modularities of all partitions (real partitions included)
We do not expect the partitions you obtain to be equal to the real ones or between them, except in certain cases.

Documentation

There is no need to elaborate a detailed documentation for these practical exercises, we just need the code, data, results and plots. You may just put the corresponding files with easy-to-identify names, distributed in folders if necessary. If you prefer to put all this information also in a single documentation file, that's OK, but avoid long texts.


----


	    igraph library:

7 community detection algorithms (including those mentionned above):
    Edgebetweenness (Girvan-Newman link centrality-based approach),
    Walktrap (Pons-Latapy random walk-based approach),
    Leading Eigenvectors (Newman's spectral approach),
    Fast Greedy (Clauset et. al modularity optimization),
    Label Propagation (Raghavan et. al),
    Louvain (Blondel et. al, modularity optimization),
    Spinglass (Reichardt-Bornholdt, modularity optimization),
    InfoMap (Rosvall-Bergstrom, compression-based approach).