https://github.com/triandicAnt/GraphEmbeddingRecommendationSystem

Python based Graph Propagation algorithm, DeepWalk to evaluate and compare preference propagation algorithms in heterogeneous information networks from user item relation ship.
https://github.com/triandicAnt/GraphEmbeddingRecommendationSystem

deepwalk graph graph-embedding graph-propagation-algorithm prediction python rating recommendation-system

Last synced: about 2 months ago
JSON representation

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

• Host: GitHub
• URL: https://github.com/triandicAnt/GraphEmbeddingRecommendationSystem
• Owner: triandicAnt
• License: mit
• Created: 2016-04-22T01:34:07.000Z (over 8 years ago)
• Default Branch: master
• Last Pushed: 2018-02-04T06:42:58.000Z (over 6 years ago)
• Last Synced: 2024-05-01T02:30:28.411Z (5 months ago)
• Topics: deepwalk, graph, graph-embedding, graph-propagation-algorithm, prediction, python, rating, recommendation-system
• Language: Python
• Size: 21.5 MB
• Stars: 189
• Watchers: 10
• Forks: 55
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• graph-networks - DeepWalk

README

        
# 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

python -m rec2vec --walk-length 2 --number-walks 2 --workers 4

# ****arguments****

# walk-length

# number-walks

# workers

```

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