Movie Recommendation System

A Full-Stack AI Web Application

This project is a movie recommendation engine that suggests films based on content similarity. It features a decoupled architecture with a FastAPI backend and a Streamlit frontend, both deployed in the cloud.

Live Links

• Frontend (UI): 
• Backend (API):  

Architecture

The system follows a Microservices Architecture:

1. Backend (FastAPI): Handles the machine learning logic, similarity scores, and TMDB API integration. Hosted on Render.
2. Frontend (Streamlit): Provides a responsive UI for users to search movies and view recommendations. Hosted on Streamlit Community Cloud.

Machine Learning Pipeline

The recommendation engine follows a sophisticated NLP workflow to understand movie context:

1. Data Engineering:
   • Cleaned a dataset of 45,000+ movies.
   • Handled complex data structures using ast.literal_eval to extract genres from stringified dictionaries.
2. Text Preprocessing (NLP):
   • Tokenization & Regex: Removed noise and punctuations.
   • Stopword Removal: Filtered non-informative English words.
   • Lemmatization: Used WordNetLemmatizer to reduce words to their root form (e.g., "running" to "run").
3. Vectorization:
   • Implemented TF-IDF (Term Frequency-Inverse Document Frequency) with an ngram_range of (1,2). This captures both individual words and meaningful pairs of words.
4. Similarity Engine:
   • Utilized Cosine Similarity to calculate the distance between movie vectors in a 50,000-dimensional space.

---

Project Structure

  ├── data/           # Raw Data
  ├── models/         # Serialized .pkl files (TF-IDF matrix, indices, model)
  ├── notebooks/      # recommendation.ipynb (Research & EDA)
  ├── src/
  │   ├── app.py      # Streamlit UI Source Code
  │   └── main.py     # FastAPI Backend Source Code
  ├── requirements.txt # Library dependencies
  └── runtime.txt      # Python version specification (3.11.9)

---

Tech Stack

• Language: Python 3.11.9
• Machine Learning: Scikit-learn (TF-IDF Vectorizer & Cosine Similarity), Pandas, NumPy, NLTK, Joblib
• Web Frameworks: FastAPI (Backend), Streamlit (Frontend)
• APIs: TMDB (The Movie Database) for posters and metadata
• Deployment: Render (Backend), Streamlit Cloud (Frontend)

---

⚙️ Setup & Installation

Step 1: Clone the Repository

git clone https://github.com/itsdakshjain/Movie-Recommendation-System
cd Movie-Recommendation-System

Step 2: Create a Virtual Environment

python -m venv venv
# On Windows:
venv\Scripts\activate
# On Mac/Linux:
source venv/bin/activate

Step 3: Install Dependencies

pip install -r requirements.txt

Step 4: Run the Application

#Run Backend
uvicorn src.main:app --reload
#Run Frontend
streamlit run src/app.py

---

License

This project is licensed under the MIT License—a permissive license that allows for personal and commercial use while providing a disclaimer of warranty. See the LICENSE file for the full text.

Contact & Credits

Developer: Daksh Jain