基于压缩-激励网络双塔模型的卷烟推荐算法

Cigarette Recommendation Algorithm Based on the SENET Dual-tower Model

采用压缩和激励网络(squeeze and excitation network,SENET)构建双塔推荐模型,针对卷烟消费推荐问题进行了研究.基于长期采集的包含用户、商品和历史交互信息的数据集,使用SENET双塔推荐模型对该数据集进行建模,以预测用户的卷烟消费行为.实验结果表明,基于SENET双塔构架的卷烟消费推荐模型在消费者与产品间的交互信息获取上具有优势;与传统的推荐算法相比,SENET双塔推荐模型具有更好的推荐效果.

This article employs the squeeze and excitation network(SENET)to construct a dual-tower recommendation model and investigates the cigarette consumption recommendation problem.Based on a long-term collected dataset containing user,product,and historical interaction information,the SENET dual-tower recommendation model is used to model the dataset and predict users′cigarette consumption behavior.The experimental results of this article show that the cigarette consumption recommendation model based on SENET dual-tower architecture has advantages in obtaining interaction information between consumers and products.Compared with traditional recommendation algorithms,the SENET dual-tower recommendation model has better recommendation performance.