摘要
针对海量评论文本的情感数据复杂性、个性化推荐领域的数据稀疏性以及冷启动问题进行研究,提出了基于改进CNN-SRBM (Convolutional Neural Network-Similarity Restricted Boltzmann Machine)文本分类的评分预测推荐研究。叙述了传统CNN文本分类以及传统RBM评分预测模型;引入改进的CNN-3C、CNN-4C文本分类模型以及改进的SRBM评分预测模型;最后,融和改进的CNN-3C模型和SRBM模型,使推荐准确率达到了95. 77%。
In This paper, the complexity of emotional data, the sparsity of data in the field of personalized recommendation and the problem of cold start is studied,and a research on rating prediction recommendation based on improved CNN-SRBM text classification is proposed. Meanwhile, the traditional Convolutional Neural Network (CNN) text classification and Restricted Boltzmann Machine (RBM) rating prediction model are described. Also, the improved CNN-3C, CNN-4C text classification models and improved SRBM score prediction models are introduced. Finally, by fusing the improved CNN-3C model and SRBM model, the recommendation accuracy rate reaches 95.77%.
作者
张倩
吴国栋
陶鸿
孙成
史明哲
ZHANG Qian;WU Guodong;TAO Hong;SUN Chen;SHI Mingzhe(Anhui Agricultural University, Hefei 230000, China)
出处
《洛阳理工学院学报(自然科学版)》
2018年第4期58-64,90,共8页
Journal of Luoyang Institute of Science and Technology:Natural Science Edition
基金
国家自然科学基金项目(31671589)