Assignment 3A: Global Baseline Estimate

Objective

The goal of this assignment is to implement a non-personalized movie recommendation system using the Global Baseline Estimate algorithm in R. The recommender will use the movie ratings dataset collected in the previous assignment and stored in PostgreSQL.

Dataset

The dataset consists of three relational tables:

• users(user_id, name)
• movies(movie_id, title, release_year)
• ratings(rating_id, user_id, movie_id, rating)

The ratings were collected through a small survey in which participants rated movies on a 1–5 scale. Missing ratings occur when users have not seen a movie. To ensure reproducibility, the GitHub repo will include SQL scripts to create and populate the PostgreSQL tables.

Algorithm Plan

I will follow the implementation steps provided in the attached spreadsheet to compute μ, user bias, and item bias (with any regularization shown).
The recommendation system will use the Global Baseline Estimate model. The workflow will include:

1. Loading ratings data from PostgreSQL into R

2. Computing the global average rating (μ)

3. Computing user bias (difference between a user’s average rating and μ)

4. Computing movie bias (difference between a movie’s average rating and μ)

5. Predicting ratings using:

   predicted_rating = μ + user_bias + movie_bias

Recommendation Strategy

For a selected user, predicted ratings will be generated for movies the user has not rated. The system will recommend the movies with the highest predicted ratings.

Anticipated Challenges

Possible challenges include small sample size, sparse ratings, and ensuring reproducibility when loading data from the database into R.