🎬🍿 CineSense – AI-Powered Movie Recommender System

An intelligent content-based movie recommendation system using NLP & Machine Learning

🚀 Project Overview

Modern OTT platforms rely on AI-powered recommendation engines to keep users engaged, CineSense AI replicates this logic using content similarity. CineSense AI is a content-based movie recommendation system built using NLP and machine learning, It processes movie metadata, cleans text, generates tags, transforms them into vectors, and uses cosine similarity to recommend movies closest in meaning.

The system performs:

• Metadata extraction
• Data cleaning
• Tag generation
• Text normalization & stemming
• Bag-of-Words vectorization
• Cosine similarity computation
• Top-5 movie recommendation

📂 Dataset Used

This project uses the TMDB 5000 Movies Dataset, containing:

• tmdb_5000_movies.csv
• tmdb_5000_credits.csv

These include rich metadata such as:

• 🎭 Genres
• 🎞️ Keywords
• 🧑‍🎤 Cast
• 🎬 Crew
• 📝 Overview
• 📊 Popularity, Ratings, Budget, Revenue

🏗️ Project Workflow

1️⃣ Importing Libraries : Used for manipulation, visualization, NLP, and model building.

• import pandas as pd
• import numpy as np
• import matplotlib.pyplot as plt
• import seaborn as sns

2️⃣ Loading the Datasets

• movies = pd.read_csv("tmdb_5000_movies.csv")
• credits = pd.read_csv("tmdb_5000_credits.csv")

3️⃣ Inspecting Data

• movies.head(2)
• movies.info()

4️⃣ Merging Movie & Credits Data

• Merged using the movie title.
• movies = movies.merge(credits, on='title')

5️⃣ Feature Selection & Cleaning Selected important fields:-

• id
• title
• overview
• genres
• keywords
• cast
• crew Removed unnecessary columns and handled missing values.

6️⃣ Converting JSON-like Strings to Lists Used ast.literal_eval to convert dictionary-like strings to Python objects.

7️⃣ Creating the tags Column Combined important columns into a single unified feature text. movies['tags'] = movies['overview'] + movies['genres'] + movies['keywords'] + movies['cast'] + movies['crew']

🧠 Text Processing & NLP

The system performs:

• Lowercasing

• Removing spaces in multi-word tokens

• Stemming using PorterStemmer

• Importing Stemmer from nltk.stem.porter import PorterStemmer ps = PorterStemmer()

• Stemming Function def stem(text): y = [] for i in text.split(): y.append(ps.stem(i)) return " ".join(y)

• Applying Stemming df['tags'] = df['tags'].apply(stem)

🔍 Testing the Stemmer

df['tags'][0], ps.stem('loved'), stem("in the 22nd century, a paraplegic marine is dispatched...")

🧮 Vectorization

• Convert textual tags into numerical vectors using Bag-of-Words.

from sklearn.feature_extraction.text import CountVectorizer cv = CountVectorizer(max_features=5000, stop_words='english') vectors = cv.fit_transform(new_df['tags']).toarray()

📐 Cosine Similarity Matrix

• Used to compute similarity between movies.

from sklearn.metrics.pairwise import cosine_similarity similarity = cosine_similarity(vectors) similarity.shape

🎯 Recommendation Function

• Returns Top 5 Similar Movies for any movie title.

def recommend(movie): movie_index = new_df[new_df['title'] == movie].index[0] distances = similarity[movie_index] movies_list = sorted(list(enumerate(distances)), reverse=True, key=lambda x: x[1])[1:6] for i in movies_list: print(new_df.iloc[i[0]].title)

💾 Model Saving

• To use the model later (deployment-ready):

import pickle pickle.dump(new_df, open('movies.pkl', 'wb')) pickle.dump(similarity, open('similarity.pkl', 'wb'))

Distributed files

File Description :

• movies.pkl Processed movie dataframe
• similarity.pkl Cosine similarity matrix
• CineSense-AI-Recommendation-System.ipynb Full Jupyter notebook
• README.md Documentation

▶️ How to Use

• Run the notebook and call:

recommend('Avatar') : Output will display the top 5 recommended movies based on content similarity.

📈 Future Enhancements

• TF-IDF vectorization
• Word2Vec / Doc2Vec embedding
• BERT semantic embeddings
• Hybrid (Content + Collaborative filtering)
• Deployment using Flask / FastAPI
• Web UI using Streamlit or React

🏁 Conclusion

• CineMatch AI successfully demonstrates how Natural Language Processing and machine learning can be applied to build a practical, content-based movie recommendation system.

• By converting movie metadata into feature vectors and measuring their similarity, the system provides accurate and meaningful movie suggestions.

• This project lays a strong foundation for more advanced recommendation engines and can be expanded with deep learning, hybrid methods, and full-stack deployment to create a production-grade system.

👨‍💻 Author

Ayush

Aspiring Data Scientist & Analyst

• 📫 Email: bhanuseenu914@gmail.com
• 🌐 GitHub: https://github.com/ayush13-0
• ℹ️ LinkedIn: www.linkedin.com/in/ayush130

🛡️ License

• This project is licensed under the MIT License.