metadata

title: Anime Recommendation System
emoji: 📚
colorFrom: purple
colorTo: yellow
sdk: streamlit
sdk_version: 1.41.1
app_file: app.py
pinned: false
license: apache-2.0

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

Anime Recommendation system

This is an Anime Recommendation System that combines multiple recommendation techniques such as Collaborative Filtering, Content-Based Filtering, and Popularity-Based Filtering. We used the AnimeList 2023 dataset, but to optimize computational cost and storage, we included only animes with an average rating above 6.0. The system is designed for continuous data ingestion and transformation and is fully dockerized for seamless deployment.

The system is hosted on Hugging Face Spaces and fetches datasets and pre-trained models from Hugging Face Hub to generate personalized anime recommendations based on user preferences and anime features.

Live Demo 🤗

Anime Recommendation System App

Tech Stacks 🛠️

Python: Main programming language used for building recommendation algorithms and Streamlit app.
Docker: Containerizes the application to ensure a consistent environment across different platforms.
Streamlit: For building and deploying the web app that serves the recommendations.
Hugging Face Spaces: Hosts the Streamlit-based recommendation system.
Hugging Face Datasets: Stores and retrieves anime datasets for processing.
Hugging Face Models: Stores the pre-trained recommendation models for inference.

Pipeline 🚀

The pipeline follows a structured sequence of steps to build an Anime Recommendation System, including data ingestion, transformation, and multiple recommendation models.

1. Data Ingestion 📥

Initiates the data ingestion process, where anime data is loaded from Hugging Face datasets.
The ingested data is saved as artifacts in a local folder for further processing.

2. Data Transformation 🔄

Cleans, transforms, and processes the raw data into a structured format.
Extracts important features required for Content-Based Filtering and prepares data for Collaborative Filtering.

3. Collaborative Filtering 🤝

Implements three collaborative filtering models to recommend anime based on user preferences:
- Singular Value Decomposition (SVD): Factorizes the user-item interaction matrix to make personalized recommendations.
- Item-Based K-Nearest Neighbors (Item-KNN): Recommends anime similar to a given anime based on user ratings.
- User-Based K-Nearest Neighbors (User-KNN): Suggests anime that users with similar preferences have liked.
The chosen model is trained using transformed data, and the final trained model is stored as an artifact.
Once trained, it can generate recommendations for users or anime titles.

4. Content-Based Filtering 🎭

Uses extracted anime features like genres to train a Content-Based Recommendation Model.
This model recommends anime similar to those a user has watched or liked.

5. Popularity-Based Filtering ⭐

This recommendation system ranks anime based on various popularity metrics, making it ideal for users who want to discover trending or highly-rated shows without needing personalized preferences.

The system applies different filters to sort anime based on:

Most Popular 🎭: Anime ranked by popularity score, highlighting the most widely recognized titles.
Top Ranked 🏆: Highest-rated anime, based on official ranking metrics.
Overall Top Rated ⭐: Best-rated anime, sorted by average user ratings.
Most Favorited ❤️: Anime with the highest number of favorites, indicating strong fan appreciation.
Highest Member Count 👥: Anime with the largest viewer base, showing widespread appeal.
Popular Among Members 🔥: Anime with a high number of members and strong ratings, making them community favorites.
Highest Average Rating 🎖️: Shows that have the best average rating after handling missing values.

Artifacts Storage 📂

All intermediate and final outputs, including processed datasets and trained models, are first saved locally in the Artifacts folder. These artifacts are then uploaded to Hugging Face for efficient storage and easy access.

When building the Streamlit app, these datasets and trained models are retrieved directly from Hugging Face, ensuring seamless integration and scalability.

The datasets used in this project are available at:
- Anime and User Ratings
You can find the Artifacts of trained models here:
- Pre-trained Models

Deployment on Hugging Face Spaces 🚀

This project is deployed using Hugging Face Spaces, which provides a seamless way to host Streamlit applications. The application pulls the datasets and trained models from Hugging Face and runs inside a Docker container.

Pre-requisites

Docker
Hugging Face (for datasets and trained models)
Python 3.8+
Hugging Face Spaces account

Local step 🔧

Clone the repository

   git clone https://huggingface.co/spaces/krishnaveni76/Anime-Recommendation-System
   cd Anime-Recommendation-System

Set Up a Virtual Environment:

# For macOS and Linux:
python3 -m venv venv 
# For Windows:
python -m venv venv

Activate the Virtual Environment:

# For macOS and Linux:
source venv/bin/activate 
# For Windows:
.\venv\Scripts\activate

Install Required Dependencies:

pip install -r requirements.txt

Running the Pipeline 🔄

To process the data and train the recommendation models, run the following command:

python run_pipeline.py

This will execute the pipeline, ingest and transform data, and train the models before making recommendations.

Running with Docker 🚀

To run the application inside a Docker container, follow these steps:

Build the Docker Image

docker build -t anime-recommendation-system .

Run the Docker Container

docker run -p 8501:8501 anime-recommendation-system

This will start the Streamlit application, which can be accessed at http://localhost:8501.

Contact 📫

For any questions, suggestions, or collaboration opportunities, feel free to reach out:

📧Email: [email protected]

🌐 LinkedIn: Krishnaveni Ponna

🐦 Twitter: @Krishnaveni076