The Effect of K - Means Clustering on Collaborative Filtering in Book Recommendation

As recorded on the Goodreads dataset, around 2 million books exist up until 2017. A recommendation system provides book recommendations based on user profiles. Collaborative Filtering (CF) is one of the methods of recommendation systems. There are several approaches, some of them are neighborhood-based, such as user-based CF and item-based CF. However, the computation is time-consuming, so clustering can be employed beforehand to create a faster model. The clustering will split users into several clusters and use CF to compute rating predictions. The results show that K-Means clustering reduces the inference time in both CF methods. Item-based CF is also found to be better suited to K-Means clustering than user-based CF.