PyDataDC Recommender System Workshop

The recommender system is a classic application of machine learning that aims to predict which item a user will like best. Personalized recommendations play an integral role for e-commerce platforms, with the goal of driving user engagement through item recommendations.

In this workshop, we will build two types of recommender systems using the MovieLens dataset:

1. an item-item recommender using k Nearest Neighbors (kNN) and cosine similarity
2. a top N recommender using matrix factorization

We will also cover the following topics on recommendations:

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

By the end of this workshop, you will have a better understanding of the different techniques and tools used to build recommendation systems in real-life scenarios.

See slides for this workshop here.

Environment Setup Instructions:

• Option 1: Running Jupyter notebook locally
• Option 2: Running Jupyter notebook via Google Colab

Recommender Basics:

• What is a recommender system?
  • Collaborative Filtering
  • Content-based Filtering

Workshop Material:

• Building a Recommender System from Scratch
  • Step 1: Import Dependencies
  • Step 2: Load Data
  • Step 3: Exploratory Data Analysis
  • Step 4: Data Pre-processing
  • Step 5: Item-item Recommendations with k-Nearest Neighbors
  • Step 6: Handling the cold-start problem
  • Step 7: Dimensionality Reduction with Matrix Factorization (advanced)
  • Step 8: Evaluation