Movie Recommendation System with NLP, TF-IDF & PCA

A content-based movie recommendation system that suggests similar movies based on genres, keywords, cast, and director using Natural Language Processing (NLP) techniques. The project includes interactive visualizations like PCA projections and word clouds to better understand relationships between movies.

---

Features

• Content-Based Recommendations (no user ratings needed)
• TF-IDF Vectorization of movie metadata
• Dimensionality Reduction using PCA for 2D visualization
• Scatter Plot of movie clusters based on content
• Word Cloud representing most frequent keywords across movies
• Clean and interpretable code with helper functions and modular structure

---

Tech Stack

Category	Technology
Programming Language	Python 3
Libraries & Tools	Pandas, NumPy, Scikit-learn
Visualization	Matplotlib, Seaborn, WordCloud
Dataset	TMDb 5000 Movie Dataset from Kaggle

---

Dataset

Dataset used: TMDb 5000 Movie Dataset

• tmdb_5000_movies.csv — Movie metadata including genres, keywords, overview
• tmdb_5000_credits.csv — Cast and crew details for each movie

---

Getting Started

1. Clone the Repository

git clone https://github.com/yourusername/movie-recommendation-system.git
cd movie-recommendation-system

2. Install Dependencies

You can install the required libraries using pip:

pip install pandas numpy matplotlib seaborn scikit-learn wordcloud

3. Download Dataset

Place the following CSV files inside the project directory:

tmdb_5000_movies.csv

tmdb_5000_credits.csv

You can download these from Kaggle.

---

Visualizations

• PCA Plot of Movies Displays a 2D scatter plot using PCA to reduce the high-dimensional TF-IDF matrix. Each point represents a movie, colored by its genre. Similar movies are closer together.

• Word Cloud The word cloud visualizes the most frequent terms from all movie metadata (overview, genres, keywords, cast, and director), helping to identify trends in the dataset.

---

How It Works

1. Data Cleaning & Preprocessing

• Parse JSON-like strings for genres, cast, crew, and keywords.
• Extract key features: top 3 cast members and director.

2. Feature Engineering

• Merge overview, genres, keywords, cast, and crew into a single tags column.
• Normalize the text: lowercase, no spaces, and stopword removal.

3. Vectorization

• Use TfidfVectorizer to convert text into feature vectors.
• Each movie is represented by a 5000-dimensional sparse vector.

4. Similarity Calculation

• Compute pairwise cosine similarity between movies.

5. PCA for Visualization

• Reduce 5000D TF-IDF vectors to 2D using Principal Component Analysis (PCA).
• Visualize the content-based proximity between movies.

6. Recommendation Function

• Takes a movie title as input.
• Returns the top 5 most similar movies based on cosine similarity.

---

Example Usage

recommend('The Dark Knight')

Top 5 Recommendations for 'The Dark Knight':
 - The Dark Knight Rises
 - Batman Returns
 - Batman Begins
 - Batman Forever
 - Batman: The Dark Knight Returns, Part 2

Learnings

• How to engineer text-based features from structured and semi-structured data.
• Implemented TF-IDF from scratch for a recommender system.
• Learned to apply PCA for visualizing high-dimensional text vectors.
• Hands-on with cosine similarity, vector math, and NLP in Python.

---

⭐️ Give it a Star

If you found this repo helpful or interesting, please consider giving it a ⭐️. It motivates me to keep learning and sharing!

---