摘要
在数字文化资源推荐中,资源与用户兴趣的精准匹配扮演着关键作用。虽然知识图谱有效地解决了传统推荐算法中数据稀疏性与冷启动问题,但知识图谱的静态结构却限制了对用户兴趣动态演化的理解。针对上述问题,文中提出了一种融合知识图谱和兴趣偏好的数字文化资源推荐方法(Knowledge Graph Interest Preferences,KGIP)。该方法首先通过构建知识图谱的嵌入表示用以建立用户与资源之间的关联关系。其次,采用长短期记忆网络模块表征用户的兴趣,并挖掘用户长短期历史行为中的复杂特征,更准确地捕捉用户的兴趣偏好。最后,为了充分利用兴趣偏好以及资源之间的关联信息,将两种特征表示进行融合送入多层感知器学习不同潜在因子之间的非线性结构特征,引入Sigmoid激活函数得到最终预测结果。通过在豆瓣平台和国家文化云平台数据集上进行多次实验验证,结果表明KGIP在数字文化资源推荐中具有良好的表现。
In digital cultural resource recommendation,the precise matching between resources and user interests plays a key role.Although knowledge graphs effectively address the data sparsity and cold start problems in traditional recommendation algorithms,the static structure of knowledge graphs limits the understanding of the dynamic evolution of user interests.To address these issues,we propose a digital cultural resource recommendation method that integrates knowledge graphs and interest preferences(Knowledge Graph Interest Preferences,KGIP).Firstly,this method establishes the association between users and resources by constructing embedding representations of knowledge graphs.Secondly,it utilizes a long short-term memory network module to characterize user interests and explores complex features in users'long and short-term historical behaviors to more accurately capture user interest preferences.Finally,to fully utilize interest preferences and the association information between resources,the two feature representations are merged and fed into a multi-layer perceptron to learn the nonlinear structural features among different latent factors,introducing the Sigmoid activation function to obtain the final prediction results.Through multiple experiments on the Douban platform and the National Cultural Cloud platform dataset,the results show that KGIP performs well in digital cultural resource recommendation.
作者
张大更
王西汉
高全力
ZHANG Da-geng;WANG Xi-han;GAO Quan-li(School of Computer Science,Xi’an Polytechnic University,Xi’an 710600,China)
出处
《计算机技术与发展》
2024年第9期124-130,共7页
Computer Technology and Development
基金
国家自然科学基金项目(61902300)。
关键词
推荐算法
知识图谱
长短期记忆网络
长短期兴趣偏好
数字文化资源推荐
recommendation algorithm
knowledge graph
long short-term memory
long short-term interest preferences
digital cultural resource recommendation
