Skip to content

chusloj/recommender-tutorial

 
 

Repository files navigation

Recommendation Systems 101

This series of tutorials explores different types of recommendation systems and their implementations. Topics include:

  • collaborative vs. content-based filtering
  • implicit vs. explicit feedback
  • handling the cold start problem
  • recommendation model evaluation

We will build various recommendation systems using data from the MovieLens database. You will need Jupyter Lab to run the notebooks for each part of this series. Alternatively, you can also use Google’s new colab platform which allows you to run a Jupyter notebook environment in the cloud. You won't need to install any local dependencies; however, you will need a gmail account.

The series is divided into 3 parts:

  1. Building an Item-Item Recommender with Collaborative Filtering
  2. Handling the Cold Start Problem with Content-based Filtering
  3. Building an Implicit Feedback Recommender System

More information on each part can be found in the descriptions below.

Part 1: Building an Item-Item Recommender with Collaborative Filtering

Description
Objective Want to know how Spotify, Amazon, and Netflix generate "similar item" recommendations for users? In this tutorial, we will build an item-item recommendation system by computing similarity using nearest neighbor techniques.
Key concepts collaborative filtering, content-based filtering, k-Nearest neighbors, cosine similarity
Requirements Python 3.6+, Jupyter Lab, numpy, pandas, matplotlib, seaborn, scikit-learn
Tutorial link Jupyter Notebook
Resources Item-item collaborative filtering, Amazon.com Recommendations, Various Implementations of Collaborative Filtering

Part 2: Handling the Cold Start Problem with Content-based Filtering

Description
Objective Collaborative filtering fails to incorporate new users who haven't rated yet and new items that don't have any ratings or reviews. This is called the cold start problem. In this tutorial, we will learn about clustering techniques that are used to tackle the cold start problem of collaborative filtering.
Requirements Python 3.6+, Jupyter Lab, numpy, pandas, matplotlib, seaborn, scikit-learn
Tutorial link Jupyter Notebook

Part 3: Building an Implicit Feedback Recommender System

Description
Objective Unlike explicit feedback (e.g., user ratings), implicit feedback infers a user's degree of preference toward an item by looking at their indirect interactions with that item. In this tutorial, we will investigate a recommender model that specifically handles implicit feedback datasets.
Requirements Python 3.6+, Jupyter Lab, numpy, pandas, implicit
Tutorial link Jupyter Notebook

About

An introduction to recommendation systems in Python

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • HTML 56.9%
  • Jupyter Notebook 42.8%
  • Python 0.3%