AI Based Book Recommendation System

---

🎯 Goal

The main goal of this project is to develop an AI-based book recommendation system that can suggest similar books to users based on a given book title. The purpose is to enhance the book discovery experience by leveraging deep learning and hybrid recommendation techniques.

🧵 Dataset

Dataset link: https://www.kaggle.com/datasets/mdhamani/goodreads-books-100k/

🎥 YouTube

🧾 Description

This project aims to improve upon an existing book recommendation system by integrating an AI model that suggests books from a large dataset. The recommendation system combines traditional content-based filtering with neural collaborative filtering (NCF) and convolutional neural network (CNN) models to enhance recommendation accuracy.

🧮 What I had done!

1. Data Preprocessing:

   • Cleaned and prepared the dataset by handling missing values and encoding categorical data.
   • Generated TF-IDF vectors for book descriptions and computed cosine similarity for content-based filtering.

2. Model Training:

   • Implemented and trained a Neural Collaborative Filtering (NCF) model using user and book embeddings.
   • Implemented and trained a Convolutional Neural Network (CNN) model to predict book ratings based on descriptions.

3. Hybrid Recommendation System:

   • Combined the results from cosine similarity, NCF, and CNN models to provide a robust recommendation system.

4. Deployment:

   • Created a Streamlit application to provide an interface for users to get book recommendations.

🚀 Models Implemented

1. Cosine Similarity: Used for content-based filtering based on book descriptions.

   • Chosen for its simplicity and effectiveness in measuring similarity between text vectors.

2. Neural Collaborative Filtering (NCF): Used for collaborative filtering based on user-book interactions.

   • Chosen for its ability to learn user and item embeddings that capture latent factors in the data.

3. Convolutional Neural Network (CNN): Used to predict book ratings based on textual descriptions.

   • Chosen for its ability to capture complex patterns in text data through hierarchical feature extraction.

📚 Libraries Needed

• pandas
• numpy
• scikit-learn
• torch
• tensorflow
• keras

📈 Performance of the Models based on the Accuracy Scores

• Neural Collaborative Filtering (NCF) Model:

  • Test MAE: 0.2378

• Convolutional Neural Network (CNN) Model:

  • Test MAE: 0.2390

• Cosine Similarity:

  • Provides top-N recommendations based on content similarity.