Genre Classification of Books Using Machine Learning Techniques

This project, titled "Genre Classification of Books Using Machine Learning Techniques," by Shaon An Nafi and Peya Pandit from Daffodil International University, aims to automate the challenging and subjective process of book genre classification. The research addressed the limitations of manual classification by applying machine learning and deep learning techniques. The methodology involved: Collecting a diverse dataset of approximately 600 book texts across six genres (Horror, Mystery, Thriller, Comedy, Romance, and Fiction). Utilizing OCR (Optical Character Recognition) for digitized physical books and extracting text from online sources. Extensive preprocessing, including text cleaning, tokenization, and TF-IDF vectorization. Training and evaluating various machine learning and deep learning models, including SVM, Random Forest, Gradient Boosting Machines, RNN, and BERT. The key finding was that the transformer-based BERT model performed the best, achieving a 75% accuracy rate, closely followed by SVM with 73% accuracy. The project culminated in a Flask-based demo web application that can take text or PDF input to predict the book's genre. This system is designed to enhance book recommendation systems, automate library management, and contribute to the preservation of cultural heritage by making literary content more discoverable and organized.