sadensmol / plan_deep_learning

Deep Learning for Beginners

This repository includes all the problems I solved during my learning of DL.

Process is split into parts related to the exact real world problem. I included some papers and lots of docs, explanations.

Environment

• Visual Studio Code

• Jupiter Notebooks

• Python 3

• Pandas

• Scikit-Learn

• Pytorch

• NumPy

• SciPy

• LightFM

• Surprise

Plan

Movie recommendation system (MRS)

The goal here - get recommendations for movies/series based on my interest in Star Wars. I will use different approaches and compare results, and ... watch recommendations!

Ralated papers

[II CF from Amazon](papers/[2003] Item-to-Item Collaborative Filtering.pdf)

How to start

Download the latest IMDB datasets from : https://datasets.imdbws.com/ and extract them into data/ folder.

Download the latest MovieLens datasets from https://grouplens.org/datasets/movielens/ and put them into data/ folder.

Solutions

The following Jupiter Notebooks contain different implementations for MRS.

Movie Recommendation Content Based Filtering

Movie recommendation system with content based filtering (CBF). It uses Cosine Similarity algorithm. Developed with pandas, numpy and scikit-learn libraries. It's using IMDB dataset.

Movie Recommendation User Based Collaborative Filtering

Movie recommendation system with user based collaborative filtering (UBCF). It uses Cosine Similarity algorithm. Developed with pandas, numpy and skikit-learn libraries. It's using MovieLens dataset, as IMDB doesn't provide user ratings, only average rating values per movie.

Movie Recommendation Item Based Collaborative Filtering

Movie recommendation system with item based collaborative filtering (IBCF). It uses Cosine Similarity algorithm. Developed with pandas, numpy and skikit-learn libraries. It's using MovieLens dataset, as IMDB doesn't provide user ratings, only average rating values per movie.

Movie Recommendation with Matrix Factorisation

Movie recommendation system with Matrix Factorisation approach. (MF). It uses SVD/SVD+ algorithms. Developed with pandas, numpy,scipy and skikit-learn libraries. It's using MovieLens 100K dataset, as IMDB doesn't provide user ratings, only average rating values per movie. I coulnd't get it working with 20M, and 25M datasets - as it overflows my memory.

Movie Recommendation with Deep Learning

Movie Recommendation with LightFM library

Movie Recommendation with Surprise library