摘要
为了解决影视资源推荐精度问题,引入隐含狄利克雷分布(LDA)完成对影视作品影评数据分析。考虑LDA无法体现各特征词重要性,将注意力机制嵌入网络,提高模型精度。结果显示,在推荐准确率测试中,以MoviesLens-1M数据进行测试,所提出模型准确率为0.936,相对同类推荐技术精度最好。由此可见,所提出推荐模型在系统稳定性、推荐效果上均有出色效果。
In order to solve the accuracy problem of film and television resources recommendation,the latent Dirichlet allocation(LDA)is introduced to complete the analysis of film and television works review data.Considering that LDA can not reflect the importance of each feature word,the attention mechanism is embedded into the network to improve model accuracy.The results show that in the recommendation accuracy test,by the MovieLens-1M data,the proposed model has an accuracy of 0.936,which is the best compared to similar technologies in terms of accuracy.From this,it can be seen that the proposed recommendation model has excellent application results in system stability and recommendation effectiveness.
作者
申菲
SHEN Fei(Department of Basic Teaching,Henan Technical Institute,Zhengzhou 450007,China)
出处
《微型电脑应用》
2024年第6期61-64,共4页
Microcomputer Applications
基金
郑州市社科联项目(1096)。
关键词
推荐算法
影视作品
LDA
注意力机制
recommendation algorithm
film and television work
LDA
attention mechanism
