sylvain-reynaud / i2cf

item-to-item collaborative filtering based on graph database

i2cf

item-to-item collaborative filtering based on graph database

i2cf is an application that uses a graph database to provide personalized recommendations to users. It is designed to help users discover items that are relevant to their interests by analyzing their preferences and similarities with other users.

The application uses the item-to-item collaborative filtering technique to make recommendations. It works by analyzing the relationships between items and identifying items that are similar to each other.

The project is built on a graph database, which makes it easy to model complex relationships between items and users. The database stores information about the items, such as their attributes and ratings, as well as information about the users, such as their preferences and behaviors.

HTTP API

Compute similarity

GET /compute_similarity_scores

Get recommendations for a user

GET /recommendations?user_id=<user_id>&limit=<limit>

Getting started

Import data to Neo4j

1. Start Neo4j with docker-compose

docker compose up -d neo4j

2. Import csv to Neo4j

csv files are in volumes/neo4j/imported_csv folder mounted to Neo4j container at /imported_csv folder:

• likes.csv
• ratings.csv
• users.csv
• videos.csv
• watchtime.csv

docker compose exec -it neo4j bash
cp /imported_csv/* /var/lib/neo4j/import/
chown -R neo4j:neo4j /var/lib/neo4j/import

Go to Neo4j browser at http://localhost:7474/browser/ and the query in utils/create_nodes_and_relationships.cypher file.

Run the application

After importing data to Neo4j, you can run the application with docker-compose:

docker-compose up -d