Let's assume you're interested in creating a recommendation engine to help recommend to users new movies based on previous ratings of movies they've watched.

There are atleast two types of recommender systems you can build:

• The idea here is to build a matrix of users (rows) vs things they've bought/watched/browsed/rated (columns)
• Compute similarity scores between users (rows)
• Find similar users to you
• Recommend stuff they've bought/watched/browsed/rated that you haven't

Downsides:

• Users can be nefarious
  • a user can spam by rating movies in a way to benefit their cause
  • Shiling Attack
• People's taste and interest change over time
• Far more users than things
  • Computationally very $$

• Find every pair of movies that were watched by the same user
• Measure similarity of their rating across all users who have watched the same movie
• Sort by movie, then by similarity strength (though there are other ways to do this)

This repository includes an implementation of an Item Based Recommender System for movies.

The dataset is available via GroupLens. It contains:

• 100,000 ratings
• 1,300 tag applications
• 9,000 movies
• 700 users
• Last updated 10/2016