Westerley/framework_event2vec

Framework para Event2Vec

Acesso para os resultados experimentais: Aqui

---

File: dataset.py

Convert the dataset csv file to a network graph.pkl using networkx. The network will be stored in datasets.

Parameters:

--edges

• edges.csv is the name of your csv file.

--sep

• \t delimiter of columns. Default parameter.

--output

• graph.pkl is the name of the network.

Example Usage

python dataset.py --edges datasets/edges.csv --output graph.pkl or

python dataset.py --edges datasets/edges.csv --sep=';' --output graph.pkl

---

File: main.py

Methods to create embeddings.

Parameters:

Obs: Default parameters are omitted.

--method

1. Node2Vec
2. DeepWalk
3. Metapath2Vec
4. LINE
5. NetMF

--input

• graph.pkl is the name of your network file.

--output

• name_embedding.csv is the name of your embedding file. Embedding will be stored in embeddings.

Example Usage

Implementation of paper node2vec: Scalable Feature Learning for Networks by Grover. A and Leskovec, J. Paper code

python main.py --method 1 --input datasets/graph.pkl --output node2vec_embedding.csv --embedding_dim 128 --p 1.5 --q 2.

Implementation of paper DeepWalk: Online Learning of Social Representations by Perozzi B, et al. Paper code

python main.py --method 2 --input datasets/graph.pkl --output deepwalk_embedding.csv --embedding_dim 128.

Implementation of paper metapath2vec: Scalable Representation Learning for Heterogeneous Networks by Dong, Y, et al. Paper code

python main.py --method 3 --input datasets/graph.pkl --output meta_embedding.csv --embedding_dim 128 --care_type 1 --epochs 100 --negative_samples 3 or

python main.py --method 3 --input datasets/graph.pkl --output meta_embedding.csv --embedding_dim 128 --care_type 0 --epochs 100 --negative_samples 3.

Implementation of paper LINE: Large-scale Information Network Embedding_ by Tang, J, et al. Paper code

python main.py --method 4 --input datasets/graph.pkl --output line_embedding.csv --embedding_dim 128 --proximity 'first-order' or

python main.py --method 4 --input datasets/graph.pkl --output line_embedding.csv --embedding_dim 128 --proximity 'second-order'.

Implementation of paper Network Embedding as Matrix Factorization: Unifying DeepWalk, LINE, PTE, and node2vec by Qiu J, et al. Paper code

python main.py --method 5 --input datasets/graph.pkl --output netmf_embedding.csv --embedding_dim 128 --small or

python main.py --method 5 --input datasets/graph.pkl --output netmf_embedding.csv --embedding_dim 128 --large --window 10 --rank 1024.