Analysis of Factors Affecting Book Ratings Using Data Mining Techniques

Abstract

The development of information technology has transformed how people interact with books, particularly through rating and reviewing on digital platforms, which generates large-scale data. Although the utilization of data mining in libraries continues to expand, research comprehensively analyzing the impact of metadata characteristics on average book ratings remains limited. This study aims to analyze the bibliographical metadata factors that influence the average rating of books using the Goodreads dataset. The method employed is a quantitative approach using data mining techniques with the random forest regression algorithm on ten thousand book records. The results indicate that the relationships between variables are non-linear, with the work text reviews count and total work ratings count being the most influential factors in predicting the average rating, while the original publication year has the lowest impact. This study concludes that reader appreciation in the digital era is heavily driven by the intensity of community interaction and discussion surrounding the book. These findings are vital as an empirical foundation for librarians in implementing data-driven decision-making for collection development and optimizing library recommendation services.