Improving Collaborative Filtering’s Rating Prediction Quality by Exploiting the Item Adoption Eagerness Information
Dionisis Margaris, Dimitris Spiliotopoulos, Costas Vassilakis · IEEE/WIC/ACM International Conference on Web Intelligence · 2019
Collaborative filtering generates recommendations tailored to the users’ preferences by exploiting item ratings registered by users. Collaborative filtering algorithms firstly find people that have rated items in a similar fashion; these people are coined as “near neighbors” and their ratings on items are combined in the recommendation generation phase to predict ratings and generate recommendations. On the other hand, people exhibit different levels of eagerness to adopt new products: according to this characteristic, there is a set of users, termed as “Early Adopters”, who are prone to start using a product or technology as soon as it becomes available, in contrast to the majority of users, who prefer to start using items once they reach maturity; this important aspect of user behavior is not taken into account by existing algorithms. In this work, we propose an algorithm that considers the eagerness shown by users to adopt products, so as to leverage the accuracy of rating prediction. The proposed algorithm is evaluated using seven popular datasets.