Movie Recommender

This project is a content-based movie recommendation system that utilizes Natural Language Processing to suggest films based on the semantic meaning of user descriptions. Unlike keyword-based search engines, this system employs deep learning embeddings to understand the context and themes provided in a text prompt.

Project Overview

The core of the application relies on converting movie metadata into high-dimensional vectors. By calculating the Cosine Similarity between a user's input and a pre-processed database of films, the engine identifies movies with the highest thematic relevance. The system is designed with a Flask backend and a modern, responsive web interface.

Dataset

The project utilizes the TMDB 10,000 Movies Dataset (Updated 2022). This dataset provides comprehensive metadata for approximately 10,000 films, including the specific fields required for both the recommendation logic and the user interface.

Methodology

The recommendation logic is divided into two primary phases:

1. Vectorization and Data Processing

The data pipeline, documented in movies_embeddings_creation.ipynb, processes the metadata through the following steps:

• Model Selection: Uses the all-mpnet-base-v2 model from the Sentence Transformers library to generate embeddings.
• Feature Extraction: Cleans and treats metadata columns for titles, overviews, and genres.
• Granular Embedding Generation: Creates separate vectors for the title, overview, and genres of each film to allow for specific weighting.

2. Recommendation Logic

The engine utilizes a weighted linear combination to determine the final similarity score:

• Weight Distribution: The system assigns $0.5$ to the overview, $0.35$ to the title, and $0.15$ to the genres.
• Similarity Calculation: Employs Scikit-Learn’s cosine_similarity to compare the input embedding against the stored vectors.
• Filtering: Automatically excludes entries with empty overviews or missing genre tags to ensure recommendation quality.

Technical Stack

• Backend: Python, Flask
• Data Science: Pandas, NumPy, Scikit-Learn, Sentence-Transformers, PyTorch
• Frontend: HTML5, CSS3 (Glassmorphism), Vanilla JavaScript

Installation and Setup

Prerequisites

• Python 3.8 or higher
• Pip package manager

Local Environment Setup

1. Clone the repository:

   git clone https://github.com/Cairo-Henrique/Movie-Recommender.git
   cd Movie-Recommender

2. Install the required dependencies
3. Generate the embedding database:
   • Open movies_embeddings_creation.ipynb and execute the cells to generate the .pkl file containing the vectors.
4. Launch the application:

   python app.py

Repository Structure

• app.py: The main Flask application handling routing and recommendation requests.
• app_functions.py: The code for the movies dataset and the main functions used in app.py.
• movies_embeddings_creation.ipynb: The research and development notebook used for data cleaning and embedding generation.
• static/: Contains the CSS for the glassmorphism interface and JavaScript for interactive UI elements.
• templates/: HTML templates for the web interface.
• movies.csv: original IMDB movies dataset.
• dataset_treatment.py: data cleaning of the original movies dataset; output: movies_fixed.py.
• movies_fixed.csv: The treated dataset containing movie metadata.