Skip to content

Latest commit

 

History

History
114 lines (65 loc) · 3.84 KB

README.md

File metadata and controls

114 lines (65 loc) · 3.84 KB
Raw

Recommender Systems - Content Based Systems and Collaborative Filtering

I) Overview

Created Recommender systems using TMDB movie dataset by leveraging the concepts of Content Based Systems and Collaborative Filtering.

II) Contents of the Jupyter Notebook

Section 1 - Exploratory Data Analysis

  • 1.1 Dropping Features
  • 1.2 Updating Features
  • 1.3 Feature Engineering

Section 2 - Simple Recommender System

  • 2.1 Creating a simple recommender system to show the top N movies at an overall level
  • 2.2 Creating a simple recommender system to show the top N movies based on genre

Section 3 - Content Based Recommender System

  • 3.1 Movie Overview Based Recommender
  • 3.2 Movie Metadata Based Recommender

Section 4 - Collaborative Filtering System

  • 4.1 Memory-Based Systems
    • 4.1.1 User-Based Collaborative Filtering
    • 4.1.2 Item-Based Collaborative Filtering
  • 4.2 Model-Based Systems

III) Key Packages/Functions used

  • pandas - 1.4.4
  • numpy - 1.21.5
  • SentenceTransformer (from sentence_transformers) - 2.2.2
    • Model used: all-mpnet-base-v2
  • cosine_similarity (from sklearn.metrics.pairwise) - 1.0.2
  • SVDpp (from surprise) - 1.1.3
  • Reader, Dataset (from surprise) - 1.1.3
  • cross_validate (from surprise.model_selection) - 1.1.3

IV) Predicted Recommendations

1) Simple Recommender System

Top 15 movies at an Overall level

image

Top 15 movies based on 'Horror' genre

image

2) Content Based Recommender System

2.1) Movie Overview Based Recommender

  • Top 15 movies recommended to a user who has watched 'The Godfather'

image

  • Top 15 movies recommended to a user who has watched 'The Notebook'

image

2.2) Movie Metadata Based Recommender

  • Top 15 movies recommended to a user who has watched 'The Godfather'

image

  • Top 15 movies recommended to a user who has watched 'The Dark Knight'

image

  • Top 15 movies recommended to a user who has watched the movies in the below movie list

movie_list = ['The Lion King', 'Se7en', 'Toy Story', 'Blade Runner', 'Quantum of Solace', 'Casino Royale', 'Skyfall']

image

3) Collaborative Filtering System

3.1) Memory-Based Systems

  • User-Based Collaborative Filtering

    • Top 15 movies recommended for user number 39

    image

  • Item-Based Collaborative Filtering

    • Top 15 movies recommended for user number 39

    image

3.2) Model-Based Systems

  • Top 15 movies recommended for user number 39

image

V) References

http://infolab.stanford.edu/~ullman/mmds/ch9.pdf

https://www.kaggle.com/code/rounakbanik/movie-recommender-systems

https://medium.com/grabngoinfo/recommendation-system-user-based-collaborative-filtering-a2e76e3e15c4

https://towardsdatascience.com/building-and-testing-recommender-systems-with-surprise-step-by-step-d4ba702ef80b

https://github.com/NicolasHug/Surprise/tree/master/examples