Audible Book Recommendation System

A machine learning-based book recommendation system that provides personalized audiobook suggestions using content-based filtering, clustering, and genre-based approaches.

🔗 Live Demo on Streamlit Cloud

Project Overview

This project analyzes 5,484 audiobooks from Audible and implements multiple recommendation algorithms to help users discover books based on their preferences.

Key Features:

• Content-based recommendations using TF-IDF and cosine similarity
• Cluster-based recommendations using K-Means (15 clusters)
• Genre-based filtering with customizable ratings
• Interactive web interface built with Streamlit

Dataset

• Source: Audible audiobook catalog
• Total Books: 5,484 (after cleaning)
• Features: Book name, author, rating, reviews, price, description, genre
• Original Files:
  • Audible_Catlog.csv (6,368 records)
  • Audible_Catlog_Advanced_Features.csv (4,464 records)

Technologies Used

• Python 3.11.9
• pandas, numpy - Data processing
• scikit-learn - Machine learning
• NLTK - Text processing
• Streamlit - Web interface
• matplotlib, seaborn - Visualizations

Installation

1. Clone the repository:

git clone https://github.com/priyanka7411/Audible_Book_Recommendation.git
cd Audible_Book_Recommendation

2. Create and activate virtual environment:

python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Download NLTK data:

python3 -c "import nltk; nltk.download('stopwords'); nltk.download('punkt'); nltk.download('wordnet')"

Usage

Run the Streamlit app:

streamlit run streamlit_app.py

The application will open in your browser at http://localhost:8501

Project Structure

├── data/                   # Dataset files
├── notebooks/              # Jupyter notebooks (EDA, cleaning, modeling)
├── models/                 # Saved ML models (.pkl files)
├── outputs/                # Visualizations and charts
├── streamlit_app.py        # Web application
├── requirements.txt        # Python dependencies
└── README.md

Methodology

1. Data Cleaning

• Removed 1,052 books with invalid ratings
• Handled missing values in reviews, prices, and descriptions
• Removed 1,265 duplicate records
• Extracted genre information from text

2. Exploratory Data Analysis

• Analyzed rating distributions, price trends, and genre popularity
• Created 7 visualizations including word clouds and correlation plots
• Identified key patterns: average rating 4.46/5, weak price-rating correlation

3. Recommendation Models

Content-Based Filtering:

• TF-IDF vectorization (3,000 features)
• Cosine similarity between book features
• Combines: title, author, genre, description

Cluster-Based:

• K-Means clustering (15 clusters)
• SVD dimensionality reduction (3000→50 features)
• Groups books by thematic similarity

Genre-Based:

• Filter books by genre
• Sort by rating and popularity
• Adjustable minimum rating threshold

Results

• Successfully processed and cleaned 5,484 books
• Built 3 different recommendation approaches
• Created interactive web application
• Generated 7 analytical visualizations
• Saved 6 trained models for deployment

Key Insights

• 60.7% of books rated 4.5 or higher
• Most popular genre: Audible Audiobooks & Originals (530 books)
• Price has minimal correlation with rating (0.06)
• Average book price: ₹952

Future Improvements

• Add collaborative filtering based on user ratings
• Deploy to cloud platform (AWS/Streamlit Cloud)
• Implement user accounts and reading history
• Add more advanced NLP techniques (BERT, transformers)
• Create REST API for integration

Important Note

The cosine_similarity.pkl file (229 MB) is excluded from GitHub due to size limits. It will be automatically regenerated from the TF-IDF matrix when you run the Streamlit app for the first time. This may take 10-20 seconds on initial load, then it will be cached.

Note: The cosine similarity matrix (229 MB) is computed automatically from the TF-IDF matrix when you first run the app. This may take 10-20 seconds initially, then it will be cached.

Author

Priyanka Malavade

• GitHub: @priyanka7411
• LinkedIn: Priyanka Malavade

Acknowledgments

• Dataset: Audible audiobook catalog
• Project guidance: Guvi Geek Networks
• Libraries: scikit-learn, pandas, Streamlit, NLTK

License

This project is for educational purposes.