|
Abstract:
|
Understanding user online behaviors plays a critical role in building recommendation systems. Most recently various neural sequence models have been proposed to depict users’ characteristics from their online activities. Unlike traditional approaches such as collaborative filtering and factorization machine models, neural sequence models exploit user’s temporal fingerprints to better understand their behaviors, resulting in a more precise recommendation framework that is able to capture user’s varying interests. In this work, we review the latest innovations of recommendation systems using neural sequence models and propose a new approach using attention based modules. Novel insights are discovered from the proposed approach.
|
