Graph-Embedding-For-Recommendation-System

Python based Graph Propagation algorithm, DeepWalk to evaluate and compare preference propagation algorithms in heterogeneous information networks from user item relation ship.

Objective:

• Predict User's preference for some items, they have not yet rated using graph based Collaborative Filtering technique, DeepWalk on user-movie rating data set.
• Firstly, using the movie review data set, a heterogeneous graph network with nodes as users, movies and its associated entities (actors, directors) were created.
• DeepWalk was used to generate a random walk over this graph.
• Theses random walks were embedded in low dimensional space using Word2Vec.
• The prediction for rating for a user-movie pair was done by finding the movie-rating node with the highest similarity to the user node.

Requirements:

• numpy
• scipy

Steps to Run:

Run the following command from root folder(not inside rec2vec)

python -m rec2vec --walk-length 2 --number-walks 2 --workers 4
# ****arguments****
# walk-length
# number-walks
# workers

Ref : https://github.com/phanein/deepwalk