Book Recommendation System

This project implements a hybrid Book Recommendation System using the Book-Crossing dataset. It suggests books to users based on their ratings and preferences by combining content-based and collaborative filtering techniques.

---

Features

• Content-Based Filtering using book metadata (title, author, publisher)
• User-Based Collaborative Filtering based on similar users' preferences
• Item-Based Collaborative Filtering for users with only one rating
• Global Top Picks fallback for users with no ratings

---

Dataset

• Source: Book-Crossing Dataset on Kaggle
• Files Used:
  • Books.csv
  • Ratings.csv
  • Users.csv

---

How It Works

1. Data Loading: Downloads and loads CSV files using kagglehub and pandas.
2. Preprocessing:
   • Fills missing values and removes zero ratings.
   • Merges Books, Ratings, and Users data.
3. Feature Engineering:
   • Combines Title, Author, and Publisher.
   • Applies TfidfVectorizer and normalizes vectors.
4. Recommendation Logic (recommend_books(user_id, ...)):
   • Chooses the appropriate method based on how many books the user has rated.

---

Recommendation Methods

1. Content-Based Filtering

• Creates a metadata profile using TF-IDF and computes similarity using cosine distance.
• Recommends books similar to those rated highly by the user (≥ 8).

2. User-Based Collaborative Filtering

• Builds a user-item matrix and finds users with similar rating behavior.
• Suggests books liked by those similar users.

3. Item-Based Collaborative Filtering

• Triggered when the user has rated exactly one book.
• Finds other users who liked the same book and recommends books they also enjoyed.

4. Global Top Picks

• Used for new users with no prior ratings.
• Recommends the highest-rated books overall.

---

Contributors

• @Khoula Adil
• @Uroosha Zehra Abidi
• @Nigarish Rehman Sarmad