基于卷积神经网络的电视频道推荐算法
2020-12-23邵文倩



摘 要: 随着网络和电视服务行业的发展,网络电视平台可以为用户提供更丰富的电视频道,但对于用户而言,挑选喜爱频道的难度大幅提高。針对此问题,本文提出了基于文本卷积神经网络(TextCNN)的电视频道推荐算法,首先根据用户历史观看行为得到隐性观看特征,以及根据用户观看电视频道不同时段的节目简介得到用户特征标签和电视频道标签,然后在卷积神经网络模型中训练得到预测评分,最后通过预测评分对目标用户进行推荐个性化推荐,同时考虑了冷启动问题,使用K-Means方法来解决。我们设计了不同推荐算法的性能对比实验,最终通过基于广电运营平台中真实数据集的实验表明我们提出的算法优于其他几种基线推荐方法,提高了推荐质量。
关键词: 电视频道;推荐算法;卷积神经网络;K-Means;TextCNN
中图分类号: TP391 文献标识码: A DOI:10.3969/j.issn.1003-6970.2020.10.039
本文著录格式:邵文倩. 基于卷积神经网络的电视频道推荐算法[J]. 软件,2020,41(10):153156
【Abstract】: Through the development of the Internet and TV service industries, Internet TV platforms can provide users with more abundant TV channels, but for users, the difficulty of choosing favorite channels has also increased significantly. To solve this problem, a TV channel recommendation algorithm based on text convolutional neural network (TextCNN) is proposed. At first, the implicit viewing characteristics are obtained according to historical viewing behavior of users, and the user characteristic labels and TV channel labels are obtained according to the program introduction of different programs of the user watching the TV channel. Then trained in the convolutional neural network model to obtain the prediction score, and finally made personalized recommendations to the target users through the prediction score. At the same time, considering the cold start problem, the k-means method was used to solve it. Performance comparison experiments of various recommendation algorithms under different evaluation indicators are designed. Finally, experiments based on real data sets in the broadcasting and television operation platform show that our proposed algorithm replaces several other optional recommendation methods and improves the recommendation quality.
【Key words】: TV channel; Recommendation algorithm; K-Means; TextCNN
0 引言
现有的大多数电视服务只提供电子节目指南(Electronic program guides,EPG),即为用户提供电视频道直播和视频点播服务,用户可以根据EPG中提供的菜单列表选择自己喜欢的节目或者频道,点播视频或者节目等。但普通的EPG并非根据用户个人偏好设计,面对琳琅满目的电视频道,如何引导用户快速找到自己喜爱的频道,从而提高用户的观看质量,成为电视服务供应商面对的至关重要的问题。
电视推荐属于推荐系统应用的一大领域,文献[1]中对电视领域推荐系统的研究综述阐明,在电视领域内很多项目的推荐已经被广泛应用。……
