Recursive LSTM with Shift Embedding for Online User-Item Interaction Prediction
Chengyu Yin, Senzhang Wang, Hao Miao · 2020
Online businesses are ubiquitous nowadays. Accurately predicting the online user-item interactions such as buying, browsing, and price comparison is critically important to many e-commerce applications including recommendation, user behavior analysis and sales forecasting. In this paper, we propose RLSTM-SE, a recursive LSTM with shift embedding model, to learn the continuously evolving embeddings of users and items for more accurately predict their dynamic interactions. We conduct preliminary evaluation on two real datasets. The results show the superior performance of the proposal over several baseline models.