Abstract
Books are divided into statically labeled genres, which serve as a springboard for how readers choose literature. However, book genres are oftentimes too broad and lack proper fine-tuning to serve as an effective recommendation system. People's preferences for books exceed narrowly-defined genres, and more knowledge to compel an alteration in a user’s behavior. Recommendation engines are a growing field in machine learning and e-commerce. New recommendation methods require further research, considering the difficulty in measuring the effectiveness of engines to determine what constitutes a ‘good’ similarity in objects. The present project utilizes a newly published comprehensive book dataset and proposes two distinct models as alternative recommendation applications. Mining and analysis methods were performed to the dataset to optimize performance of predictions. In addition, visualizations of data and clusters are included to demonstrate the extracted knowledge and patterns.