Movie Recommender System 🎬

A small project built to practice pandas, SQL, and git by putting together a movie recommendation system on top of the MovieLens dataset.

How it works

1. Load & explore — checking_data.py loads the raw MovieLens CSVs with pandas, checks for nulls/duplicates, and explores the data: highest-rated movies, most active users, genre distribution, and the correlation between release year and rating.
2. Build the database — schema.py creates a SQLite database with movies, ratings, users, and tags tables, then loads the cleaned CSVs into it (extracting release year out of the movie title along the way, and pre-computing each user's rating count/average).
3. Recommend — Main.py is the actual recommender: it asks for a movie you liked, then queries the database for the recommendation logic below.

Recommendation logic

1. You type in a movie title (partial matches work — typing "Toy Story" will catch all the sequels too).
2. The script finds every user who rated that movie 4 stars or higher.
3. It looks at what else those users rated 4+ stars.
4. It returns the top 10 of those movies, ranked by average rating, as your recommendations.

In short: people who liked this also liked these.

Dataset

Built on the MovieLens 100K dataset from GroupLens Research — about 100,000 ratings across thousands of movies. The raw CSVs and the generated SQLite database aren't included in this repo (see .gitignore); download the dataset yourself to run it.

Running it

1. Download the MovieLens 100K dataset and place movies.csv, ratings.csv, and tags.csv somewhere accessible.
2. Update the file paths in schema.py (and checking_data.py, if you want the EDA too) to point to your dataset and where you want the SQLite database created — they're currently hardcoded to a local path.
3. Run the schema script once to build the database:

   python scripts/schema.py

4. Get recommendations:

   python scripts/Main.py

   What movie did you enjoy recently: Toy Story
   Movies matching your search:
   Toy Story (1995)
   Toy Story 2 (1999)
   ...
   Because you enjoyed these, you might also like:
   ...
   

What's next

• Parameterize file paths (config file or CLI args) instead of hardcoded paths
• Move the recommendation query into a proper .sql file rather than an f-string
• Weight recommendations by number of ratings, not just average rating, to avoid niche movies with one 5-star rating outranking genuinely popular ones