A Recurrent Model with Self-attention for Product Repurchase Recommendation
Pengda Chen, Jian Li · 2019
With the further development of e-commerce, new retail platforms have emerged. In the new retail platform, users have obvious cyclical characteristics for the purchase of fresh goods. In order to improve the efficiency of consumers' shopping, it is necessary to model user periodic purchase rules and recommend products for repurchase. At present, most of the recommendation models for product repurchase are based on feature engineering. These models have many problems, such as large amount of engineering, incomplete feature extraction and so on. In this work, we first propose a product repurchase recommendation model based on the Long Short Term Memory Network, which can automatically mine the temporal information in the user's shopping records to model the user periodic purchase rules. To capture the evolution of the user periodic purchase rules, we further introduce self-attention into the LSTM-based model and propose the AttRNN model. Detailed experiments on Instacart datasets are carried out to verify the effectiveness of AttRNN.