🍭Classifies Complex Networks given a list of network features. 🍭
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
MLNet-2.0
.gitignore
.travis.yml
LICENSE
README.md
requirements.txt

README.md

🍭 MLNet - Classifying Complex Networks with ML 🍭

Summary

The study of complex networks pervades all of the science. We can assign complex networks into four general classes (although there is some overlap between them): technological networks (e.g., Internet, the telephone network, power grids, transportation network), information networks (e.g., the world wide web, citation networks), biological networks (e.g., biochemical network, neutral networks, ecological networks), and social networks.

Characterizing complex network's structure is a key to understand any unifying principles underlying their topology. Several previous works have shown that many topological properties can vary for different types of system. However these works generally focus only a few characteristics at time. In this project, we present the first part of a method to characterize complex networks by performing an extensive analysis of the global and local topological features of networks. In a second part, these features are used into input vectors for a SVN classifier, establishing an efficient way of learning the classification of complex networks.

Whitepaper with results is available here.

Data

To use this software you can extract the data and calculate the features with this repository.

And cleanse the data with this repository.


Features

The feature vectors were extracting using MNet in this repository.

We have vectors for different normalization (Snowball and Metropolis Hastings Random Walk samplings) for different sizes. We also have vectors for the entire graphs for some of the features (that were possible to be calculated).

These vectors are parsed and cleansed using this repository.


Feature Section and Classifiers

We perform classification of the network features using many classifiers:

  • SVM (supervised)
  • Logistic Regression (supervised)
  • Adaboost (supervised)
  • EM (unsupervised)

Analysis and Plots

The comparisons of the the many classifiers and the plots are available under each classifier's folder.


Installation

$ pip install -r requirements.txt

License

When making a reference to my work, please use my website.

Creative Commons License

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.