A COMPREHENSIVE SURVEY ON ADAPTIVE BOOK RECOMMENDATION SYSTEM USING DEEP NEURAL NETWORKS AND MACHINE LEARNING FUSION

Abstract

The rapid growth of digital libraries and online reading platforms has significantly increased the demand for intelligent and personalized book recommendation systems that can effectively assist users in discovering relevant reading materials. However, conventional recommendation techniques often rely on limited metadata, collaborative filtering, or shallow semantic representations, resulting in reduced recommendation accuracy and limited adaptability to evolving user preferences. The proposed framework utilizes the Goodreads Books Metadata and User Interactions datasets, containing book information, user ratings, interaction records, author details, genres, and descriptive text collected from the Goodreads platform. Data preprocessing includes missing value handling, duplicate removal, categorical encoding, metadata engineering, text cleaning, feature normalization, and semantic embedding generation using a pre-trained Bidirectional Encoder Representations from Transformers (BERT) model. The extracted semantic embeddings are integrated with user behavioral and book interaction features to construct a comprehensive feature representation. Extreme Gradient Boosting (XGBoost), CatBoost, and Light Gradient Boosting Machine (LightGBM) are implemented as base regression models, while a Weighted Voting ensemble combines their predictions for personalized rating estimation and Top-N recommendation generation. Performance is evaluated using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R² Score. The proposed Weighted Voting ensemble achieves the best performance with an MSE of 0.860, RMSE of 0.927, MAE of 0.462, and an R² Score of 0.809, demonstrating improved recommendation accuracy, adaptability, and scalability for personalized book discovery. Index Terms: Book recommendation systems, collaborative filtering, content-based recommendation, hybrid recommendation, deep learning, transformer models, machine learning, personalized recommendation.