A movie recommender on movie-lens data base served with a Flask-restful app
This recommendation system uses the concepts of both memory and model based filtering to make top N recommendations.
The Jupyter Notebook shows a step by step
walkthrough. The steps in the notebook have been rearranged in the three scripts: Build_model.py, model.py and
app.py to facilitate serving the recommender in service.
You can make different types of requests which will take you to different endpoints.
In the directory of Recommender-movie, run the API locally: > python app.py.
Using curl in the terminal for example, you can make a GET request at the URL of the API to get top n movies, similar to a particular movie
via content, collaborative or hybrid filtering:
curl -X GET http://127.0.0.1:8000/movies/[basis] -d movie=[movie] -d limit=n
where [basis] can be content, collaborative or hybrid, [movie] is the name of a movie in the data base (it is usually followed by the year of production), n is an integer. For more information see the article.
ex:
> curl -X GET http://127.0.0.1:8000/movies/collaborative -d movie="Inception (2010)" -d limit=10
Then you will get a list:
{
"collaborative": [
"Shutter Island (2010)",
"Inglourious Basterds (2009)",
"Avatar (2009)",
"Social Network, The (2010)",
"District 9 (2009)",
"Black Swan (2010)",
"Sherlock Holmes (2009)",
"Kick-Ass (2010)",
"King's Speech, The (2010)",
"Up (2009)"
]
}
Another form of recommendation can be made for a particular user with an id: [userId] to get a list of n movies, that this user has not yet rated, but will likely rate highly. The curl request below will recommend a top 10 list to user=77:
> curl -X GET http://127.0.0.1:8000/users/77 -d limit=10
the result:
[
"Cosmos (1980)",
"Louis C.K.: Hilarious (2010)",
"Louis C.K.: Shameless (2007)",
"From the Earth to the Moon (1998)",
"Harakiri (Seppuku) (1962)",
"Godfather, The (1972)",
"Seven Samurai (Shichinin no samurai) (1954)",
"Stalker (1979)",
"Frozen Planet (2011)",
"Persona (1966)"
]