/Recommender-Systems-with-Collaborative-Filtering-and-Deep-Learning-Techniques

Implemented User Based and Item based Recommendation System along with state of the art Deep Learning Techniques

Primary LanguageJupyter NotebookMIT LicenseMIT

Recommender-Systems-with-CF-and-DL-Techniques

In this repository, I have covered following topics -

  • What are Recommendations Systems?
  • Why do we need Recommendation Systems?
  • Collaborative Filtering
  • Types of Collaborative Filtering
  • Memory Based CF
  • User-Based CF
  • Item-Based CF
  • Model Based CF
  • K-Nearest Neighbours
  • Singular Value Decomposition
  • Non-Negative Matrix Factorization
  • Matrix Factorization using Deep Learning
  • Introduction to Embedding Layer
  • Architecture 1 with dot operation
  • Architecture 2 with concatenation operation
  • Evaluating RMSE
  • References

You can find the kernel on Kaggle too - Recommender Systems with CF and DL Techniques

I have used Movielens 100k ratings dataset to study about various Recommendation Techniques. Since the dataset size is small, I have used basic techniques but with more size we need to use hybrid and dimensionality reduction techniques.

I have covered one such recommendation technique using autoencoder in another repository (here). This is currently the second best recommendation technique, released by NVIDIA in 2017 named Training Deep AutoEncoders for Collaborative Filtering.