Preparation for the model running

Before running the code, download the diversity file from the Baidu Cloud
Link: https://pan.baidu.com/s/1k23eRsMqoctRheuA13gTIw Password: c9jf

Unzip the DiversityFiles into AcrossNetworkEmbeddingData dir.

For instance "AcrossNetworkEmbeddingData/DiversityFiles/all_share_diversity.9.foldtrain.number"

Model Runing

All the model files in FinalModel package.

Run IONE.java and IONEDiversity.java first, then several embedding file will be generated in foursquare/embeddings and twitter/embeddings.

For IONE.java 4 embeddings,

foursquare/twitter.embedding.update.2SameAnchor.concatenate.100_dim.9.foldtrain.10000000 IONE-Ex model foursquare/twitter.embedding.update.2SameAnchor.InputContextVector.100_dim.9.foldtrain.10000000 foursquare/twitter.embedding.update.2SameAnchor.OutputContextVector.100_dim.9.foldtrain.10000000 foursquare/twitter.embedding.update.2SameAnchor.twodirectionContext.100_dim.9.foldtrain.10000000 IONE model

For IONEDiversity.java 4 embeddings,

foursquare/twitter.embedding.update.2SameAnchor.concatenate.100_dim.9.foldtrain.diversity.100000000 IONE-D-Ex model foursquare/twitter.embedding.update.2SameAnchor.InputContextVector.100_dim.9.foldtrain.diversity.100000000 foursquare/twitter.embedding.update.2SameAnchor.OutputContextVector.100_dim.9.foldtrain.diversity.100000000 foursquare/twitter.embedding.update.2SameAnchor.twodirectionContext.100_dim.9.foldtrain.diversity.100000000 IONE-D model

Then run ConcatenateAnswer.java for concatenating the embeddings,

foursquare/twitter.embedding.update.2SameAnchor.concatenateDiversity.100_dim.9.foldtrain.diversity.100000000 IONE-Con-Ex model

Of course, you can change the paths of different embeddings files for different models which are described in the paper.

Finally, you can run the getPrecision.java for the results which are used in the paper. Just make sure that the path of embedding files to be concatenated are corrected.

Data description

Description of Data directory

AcrossNetworkEmbeddingData

foursquare:

	following: the relation file, "1  2" means user 1 is the follower of user 2.  			   			

twitter:

	the same as the foursquare fold
	
twitter_foursquare_groundtruth:

	groundtruth: the groundtruth for our experiment, the anchor users between twitter and foursquare. Only column in which ids are appeared in both foursquare/following and twitter/following. 

	Note: 
	
	**pls make the anchors as the *same* id during the pre-preparation.**
	
	**Although the testing anchors have the same id, they will *not* take part in the training progress as they are not contained in the groundtruth.x.foldtrain.train file.**
	
	**There are several users have same ids (except for anchor users) in both foursquare/following and twitter/following files, but they are not the same user.**
	
	groundtruth.x.foldtrain.train, the traning anchors, which are the 0.x of all the anchors.
	
	groundtruth.x.foldtrain.test,  the testing anchors, which are the 1-0.x of all the anchors.
	
DiversityFiles

	all_share_diversity.x.foldtrain.number, which are the 0.x of all the anchors. format:
	
	user_in_networkx user_in_networky|shared_anchor_user|diversity_value|relationship_type(follower/followee)

Using other datasets for this model

Prepare your data format the same as twitter/following, foursquare/following and the twitter_foursquare_groundtruth.

Generate the DiversityFiles

Run getSharedFileAllDiversity.py in AcrossNetworkEmbeddingData dir, make sure that you have the correct path for saving generate diversity files, eg. AcrossNetworkEmbeddingData/DiversityFiles/.

Note that, use py 2.7 for the enviornment, pip install python-louvain for community detection. Then you can run the model (The same as model running section).

By the way, the you can find the IONE model at https://github.com/ColaLL/IONE and ABNE model https://github.com/ColaLL/ABNE which are also used for the user alignment task.

Feel free to contact me (Liu Li [email protected]) when you have any problems about the paper or the code.