Movie Recommendation System (IIITD PreCog'19)
The project explores three different algorithms to recommend K movies provided that user has already rated some movies. The dataset used has been downloaded from (http://files.grouplens.org/datasets/movielens/ml-latest-small.zip) which contained movies along with ratings according to different users. A web scraping script has been used to scrape about 780 movies from IMDB containing particulars about title, year of release, thumbnail, IMDB rating and Synopsis.
Getting Started
Install all the requirements using
pip install requirements.txt
Run the application using
python app.py
File Structure
- Web based application using Heroku by User Based collaborative filtering (https://movierecomm388.herokuapp.com/?fbclid=IwAR2V_f2cSfglqHSVNx3-pM9yrdVEju64SJ-AgyOml0WD1p2Q3x0uYrN0CFM).
- The ipynb notebooks for Matrix Factorization Method and Item based collaborative filtering algorithms has been uploaded in Notebooks directory.
- The web application was only deployed for the standout performing technique of User Based collaborative filtering.
Execution
The data was scraped from IMDB website and uploaded on mLab and the whole execution was carried out using Heroku.
Dataset
The dataset used is a subsample of original Movielens dataset containing data of about 780 movies. The datafiles include-
- IIITDPreCog_movies - Title of movies mapped with movieId.
- IIITDPreCog_ratings - movieId mapped with user ratings.
- IIITDPreCog_IMDB_scraped - Scraped data from IMDB website containing Title,year of release, IMDB rating, thumbnail and Synopsis.