🎥 Movie Recommendation System (Content-Based)

📌 Project Overview

Movie Recommendation System This project is a content-based movie recommendation system that suggests movies similar to a given movie. It leverages movie metadata (such as genres, keywords, cast, crew, and overview) to make accurate recommendations. The system is built using Python, Pandas, Scikit-learn, and Streamlit for deployment, providing an interactive web interface for easy usage.

---

📌 Features

• ✅ Data Preprocessing: Handles missing values and extracts relevant features like genres, keywords, and cast.
• ✅ Exploratory Data Analysis (EDA): Visualizes key patterns and insights in the movie dataset.
• ✅ Feature Engineering: Transforms data into suitable formats for machine learning.
• ✅ Machine Learning Model: Implements cosine similarity between TF-IDF vectors for movie recommendations.
• ✅ Web App with Streamlit: Allows users to input a movie title and get recommendations based on similarity.

---

📂 Project Structure

MovieRecommendationSystem/
│── dataset/
│   ├── tmdb_5000_movies.csv          # Movie metadata
│   ├── tmdb_5000_credits.csv         # Cast and crew data
│── app.py                            # Streamlit app for prediction
│── movie_recommendation.py           # Logic for movie recommendation
│── movies.pkl                        # Serialized movie data (used for prediction)
│── movies_dict.pkl                   # Serialized movie metadata dictionary
│── requirements.txt                  # Required Python libraries
│── README.md                         # Project documentation

• ✅ dataset/: Contains movie metadata and cast information.
• ✅ app.py: Streamlit app for generating movie recommendations.
• ✅ movie_recommendation.py: Core recommendation logic.
• ✅ movies.pkl: Serialized movie data for quick access.
• ✅ movies_dict.pkl: Serialized movie metadata.
• ✅ requirements.txt: List of dependencies for the project.
• ✅ README.md: Project documentation.

---

📁 Dataset

The dataset used includes two CSV files from Kaggle:

• movies.csv – Basic movie info
• credits.csv – Cast and crew details

---

📊 Data Preprocessing & EDA

The dataset contains:

• movieId (Unique ID)
• title (Movie title)
• genres (Movie genres)
• keywords (Related keywords)
• cast (Movie cast)
• crew (Director, producers, etc.)
• overview (Movie description)

🔹 Handling Missing Data

Keywords: Filled with empty strings where missing. Overview: Left untouched, as missing values indicate that no description is available.

🔹 Encoding Categorical Variables

Genres: One-hot encoding for different genres. Keywords: Vectorized using TF-IDF for textual information. Cast and Crew: Extracted and cleaned for use in the recommendation model.

---

🔍 Model Training

Algorithm Used: Cosine Similarity with TF-IDF Vectorization

from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity

tfidf = TfidfVectorizer(stop_words='english')
tfidf_matrix = tfidf.fit_transform(movies['overview'])
cosine_sim = cosine_similarity(tfidf_matrix, tfidf_matrix)

---

🧠 How It Works

1. Preprocessing

   • Merge movies.csv and credits.csv
   • Clean and combine relevant features: title, genres, overview, keywords, cast, crew
   • Convert text data into lowercase and remove spaces/special characters

2. Feature Engineering

   • Create a new feature called tags which combines:
     • Overview
     • Genres
     • Keywords
     • Cast (top 3)
     • Director (only one)

3. Vectorization

   • Use CountVectorizer from scikit-learn to convert tags into numeric vectors

4. Similarity Calculation

   • Compute similarity using cosine similarity
   • Recommend top N movies with the highest similarity score

---

Model Performance

• The model provides accurate movie recommendations based on content similarity.
• Cosine similarity calculates similarity between movies, considering their metadata (genres, keywords, etc.).

---

🖥️ Running the Web App

Run Streamlit App

streamlit run app.py

Alternative: Run via Python Script

python -m streamlit run app.py

Access in Browser Once running, open in your browser.

---

🛠️ Future Improvements

• ✅ Implement collaborative filtering to enhance recommendations with user ratings.
• ✅ Add more machine learning models like Random Forest or SVM for better predictions.
• ✅ Deploy the app using Heroku or AWS for public access.

---

🤝 Contributing

Feel free to contribute by submitting a pull request or reporting issues!

---

📩 Contact

📧 Email: ramnrngupta@gmail.com 📌 GitHub: ram-narayan-gupta-02