摘要
在电信网、广播电视网、互联网三网融合的背景下,越来越多的家庭对机顶盒电视有了更高的需求。本文基于用户收视信息、产品信息,利用文本卷积神经网络进行电视产品名称的精确提取,排除了集数、时间等对电视产品名称提取的干扰;利用改进的协同过滤算法,从付费意愿等七方面确定用户特征向量,采用余弦相似度计算用户相似度矩阵以实现相似推荐,再从电视节目类型、地区、语种、主要演员四方面为电视产品贴标签,根据相同电视产品相似度实现预期型推荐。实验结果显示,本研究成果实现了相似用户的精准推荐以及预期型推荐,可推广至网络视频推荐领域。
Under the background of the integration of telecommunication network,broadcast television network and Internet,more and more families have a higher demand for STB TV.Based on the user's viewing information and product information,this study uses text convolution neural network to extract TV product name accurately.The interference of the number of sets and time on the extraction of TV product name is eliminated;the improved collaborative filtering algorithm is used to determine the user feature vector from seven aspects such as willingness to pay,and the cosine similarity matrix is used to calculate the user similarity matrix to achieve similar recommendation.Then the TV product is labeled from four aspects such as TV program type,region,language and main actors,according to the similarity of the same TV product to achieve the expected recommendation.Finally,the experimental results show that the accurate recommendation and expected recommendation of similar users are realized,and can be extended to the field of network video recommendation.
作者
章胤
耿燕
李佳霖
李世琪
王旭
陈昭名
ZHANG Yin;GENG Yan;LI Jia-lin;LI Shi-qi;WANG Xu;CHEN Zhao-ming(School of Science,Yanshan University,Qinhuangdao 066004,China)
出处
《长春师范大学学报》
2021年第2期77-81,共5页
Journal of Changchun Normal University
基金
教育部产学合作协同育人项目“基于卷积神经网络与改进的协同过滤的电视产品推荐研究”(S201910216069)。
关键词
电视产品推荐
卷积神经网络
协同过滤
智能推荐模型
TV product recommendation
convolutional neural network
collaborative filtering
intelligent recommendation model
