| Metadata-Version: 2.1 | |
| Name: recommendationSystem | |
| Version: 0.0.1 | |
| Summary: Recommendation system + RAG Chatbot | |
| Home-page: https://github.com/Zenith40/recommendation-system | |
| Author: Satkar | |
| Author-email: [email protected] | |
| Project-URL: Bug Tracker, https://github.com/Zenith40/recommendation-system/issues | |
| Description-Content-Type: text/markdown | |
| License-File: LICENSE | |
| # 🎌 Anime Recommendation System | |
| This project is an anime recommendation engine that suggests similar anime based on the storyline using **cosine similarity**. It processes and analyzes a dataset of over **7,000+ anime** titles collected through web scraping, and applies **natural language processing (NLP)** techniques to generate meaningful recommendations. | |
| The application is containerized using **Docker** for consistent deployment across environments. | |
| --- | |
| ## 🚀 Features | |
| - Recommends animes based on **storyline similarity** | |
| - Built using **CountVectorizer + Porter Stemmer + Cosine Similarity** | |
| - Cleaned and preprocessed data pipeline from raw scraped sources | |
| - Fast, scalable recommendation system | |
| - Dockerized for smooth deployment | |
| --- | |
| ## 🛠 Tech Stack | |
| - **Python 3.9+** | |
| - **scikit-learn** for CountVectorizer & Cosine Similarity | |
| - **NLTK** for text preprocessing (PorterStemmer) | |
| - **Flask/FastAPI** (in `app.py`) | |
| - **Docker** for containerization | |
| - **BeautifulSoup / Scrapy** (for scraping, if applicable) | |
| --- | |
| ## 📊 Methodology | |
| 1. **Data Collection** | |
| - Scraped storyline and metadata for 12,000+ animes. | |
| - Structured data in a tabular format and removed duplicates, nulls, and irrelevant entries. | |
| 2. **Text Preprocessing** | |
| - Converted to lowercase | |
| - Removed special characters and stop words | |
| - Applied **Porter Stemming** to normalize tokens | |
| 3. **Feature Extraction** | |
| - Used **CountVectorizer** to convert text into token count vectors | |
| - Removed high-dimensional noise with feature filtering | |
| 4. **Similarity Calculation** | |
| - Used **Cosine Similarity** to compute similarity scores between anime storylines | |
| 5. **Recommendation** | |
| - Returned the top N animes with the highest similarity to a selected title | |
| --- | |
| ## 📦 Docker Deployment | |
| * To run the app using Docker: https://hub.docker.com/r/zenith40/recommendation-system | |
| ### Pull the Image | |
| * docker pull zenith40/recommendation-system:Ver-0.0.1 | |
| ### Run the container | |
| * docker run -p 8501:8501 zenith40/recommendation-system:Ver-0.0.1 | |
| * Then, open your browser and navigate to http://localhost:8501 | |
| --- | |
| ## 🌱 Future Improvements | |
| 🔍 Integrate TF-IDF or Word2Vec/Doc2Vec for deeper context awareness | |
| 🧠 Experiment with transformer-based models (e.g., BERT embeddings) | |
| 🌐 Deploy with a front-end UI for better user interaction | |
| ⚡ Implement caching for faster response times on repeated queries | |
| 📱 Build a mobile app version with React Native or Flutter | |
| 📊 Add filters based on genre, year, popularity, or user ratings | |
| 👥 Integrate with collaborative filtering or hybrid recommendation systems | |
| 📄 License | |
| This project is licensed under the MIT License | |
| 🙌 Acknowledgements | |
| * NLTK | |
| * scikit-learn | |