融合注意力机制的深度学习推荐模型  

Deep learning recommendation model fused with multi-source information

作　　者：于蒙 蔡利平[1] 周绪川[1] 戴涵宇 YU Meng;CAI Li-ping;ZHOU Xu-chuan;DAI Han-yu(The Key Laboratory for Computer Systems of State Ethnic Affairs Commission,Southwest Minzu University,Chengdu 610041,China)

机构地区：[1]西南民族大学计算机系统国家民委重点实验室,四川成都610041

出　　处：《西南民族大学学报（自然科学版）》2022年第5期550-560,共11页Journal of Southwest Minzu University(Natural Science Edition)

基　　金：中央高校基本科研业务费专项基金(2020NYB41);四川省科技项目(2022NSFSC0530);四川省中医药科研专项(2021ZD017)。

摘　　要：推荐算法的重要应用场景之一是电影推荐,现有的多数推荐模型利用单独的辅助信息进行推荐,一定程度上缓解了推荐不准确问题,有效利用多源信息是提升推荐效果的方式之一.设计了融入注意力机制的残差网络(ARN)模型提取电影海报的特征,增强神经网络对于局部重点区域的判别,从而对电影海报影响力大的区域权重进行调整;将海报不同类型的多源特征信息作为推荐模型的输入,提出了一种融合多源信息的深度学习推荐模型.最后,通过与多种模型在不同数据集上的性能指标进行对比,验证模型的有效性,且能够缓解数据解稀疏性问题,提升推荐模型的推荐性能.One of the important application scenarios of recommendation algorithm is movie recommendation.Most existing recommendation models use independent auxiliary information to recommend,which alleviates the inaccurate recommendation problem to a certain extent.Effective use of multi-source information is one of the ways to improve the recommendation effect.A residual network(ARN)model incorporating the attention mechanism was designed to extract the features of movie posters and enhance the discrimination of local key regions by neural network,so as to adjust the weight of the regions with high influence of movie posters.Taking the multi-source feature information of different types of posters as the input of recommendation model,a deep learning recommendation model integrating multi-source information was proposed.Finally,the effectiveness of the model was verified by comparing with the performance indicators of various models in different data sets,and the sparsity problem can be solved to improve the recommendation performance of the recommendation model.

关 键 词：深度学习 残差网络模型 注意力机制 推荐算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]