🎬 Movie Recommendation App

A Streamlit-based web app that recommends movies based on your selected title using item-based collaborative filtering.

Running the App

Option 1: Streamlit (Local)

pip install -r requirements.txt
streamlit run app.py

Option 2: Run with Docker

docker pull pavansingu/streamlit-rec-app:latest
docker run -p 8501:8501 pavansingu/streamlit-rec-app:latest

Features

• Recommend top 5 similar movies based on cosine similarity.
• Interactive UI built with Streamlit.
• Movie poster images fetched using DuckDuckGo Search (ddgs).
• Pretrained KNN model for fast recommendations.
• Docker support for easy deployment.

Working

• Input data from movies.csv and ratings.csv is used to build a user-item matrix.
• A pivot table is created to represent users' ratings for movies.
• Cosine similarity is calculated between items (movies).
• A K-Nearest Neighbors model (NearestNeighbors from sklearn) is trained on this matrix.
• When a movie is selected, the app finds and displays 5 most similar movies.
• Poster images are fetched using the ddgs (DuckDuckGo Search) package.

Input Files

• movies.csv — Metadata about each movie (e.g., movieId, title).
• ratings.csv — User ratings for movies (userId, movieId, rating).
• knn_recsy_model.pkl: Pretrained KNN model using cosine similarity.
• A correlation matrix using cosine distance is computed and used during recommendation.