Implementation of community-based graph embedding for user classification.

• Graph-based, multi-label user classification experiment demo.
• Implementation of the ARCTE (Absorbing Regularized Commute Times Embedding) algorithm for graph-based feature extraction.
• Implementation of other feature extraction methods for graphs (Laplacian Eigenmaps, Louvain, MROC).
• Evaluation score and time benchmarks.

• numpy
• scipy
• h5py
• scikit-learn
• Cython
• networkx
• python-louvain

To install for all users on Unix/Linux:

python3.4 setup.py build
sudo python3.4 setup.py install

• SNOW2014Graph dataset: Included anonymized in this project.
• The Arizona State University social computing data repository contains the ASU-Flickr and ASU-YouTube datasets.
• The Insight Project Resources repository contains the Multiview datasets in which the PoliticsUK dataset can be found.

• Implemented methods: ARCTE, BaseComm, LapEig, RepEig, Louvain, MROC.
• Other methods' implementations: LINE, DeepWalk, [EdgeCluster](http://leitang.net/social dimension.html), RWModMax, BigClam, OSLOM.

• Link

• Follow instructions on file: reveal_graph_embedding/experiments/demo.py

• If you installed the package, you will have an installed script called arcte.
• The source is located in reveal_graph_embedding/entry_points/arcte.py

If you find this code useful and use it in your research, please acknowledge its use and cite the following paper: Rizos, G., Papadopoulos, S., & Kompatsiaris, Y. (2017). Multilabel user classification using the community structure of online networks. PloS one, 12(3), e0173347.