Because of the need of scientific research, I had the honor to read CLF and tried to reproduce the paper. This Repository is a trial of reproduction of the paper:
CLF: Zhang J , Yu P S . Integrated anchor and social link predictions across social networks[C] International Conference on Artificial Intelligence. AAAI Press, 2015.
The CLF can either predict social links among users in another network as well as anchor links aligning two networks.
Before to execute CLF, you should install the following packages:
pip install networkx
pip install sklearn
The version of python is python==3.7.2 and networkx==2.2, sklearn==0.20.3, but they are not mandatory unless it doesn't work.
We provide a DBLP dataset and its distributed copy in ./graph/ called G1 and G2 which are extracted from [Prado et al., 2013] to show the effect of CLF. The data are named as DBLP1.edges and DBLP2.edges respectively, in which each line consists of node ui and node uj within one network:
ui,uj
In addition, the ground truth alignments are also needed to compute the alignment accuracy. The file is named as DBLP.alignment in ./alignment/, in which each line consists of node ui in G1 and node vi in G2:
ui,vi
In order to run CLF on the DBLP & distributed copy, execute the following command in ./src/:
python main.py --filename DBLP
If you need to modify the parameters, the complete execution command is (The parameters are optimal ones given in the paper):
python main.py --filename DBLP --align_train_prop 0.2 --alpha1 0.6 --alpha2 0.6 --c 0.1
You can check out the other options available to use with CLF using:
python main.py --help
In order to evaluate the effect of anchor link prediction, we use Top@30 and AUC to show the results.
If there are some factual errors, please let me know.
[1] Adriana Prado, Marc Plantevit, Celine Robardet, and J. F. Boulicaut. Mining graph topological patterns: Finding covariations among vertex descriptors. IEEE Transactions on Knowledge & Data Engineering, 25(9):2090–2104, 2013.
[2] Zhang J , Yu P S . Integrated anchor and social link predictions across social networks[C] International Conference on Artificial Intelligence. AAAI Press, 2015.