摘要
传统的推荐算法面临数据稀疏、泛化能力不强等问题,导致推荐准确率不高.近10年来深度学习由于较强的表达能力以及灵活的模型结构成为了最热门的研究方向.该文提出一种基于读者兴趣挖掘的深度学习推荐模型(Deep Reader Preference Ming For Book Recommendation,DRPM),该算法运用自然语言处理技术提取图书丰富的语义特征,利用长短时记忆网络模型对读者历史借阅记录进行建模,分析读者兴趣,并引入注意力机制使图书与读者进行有效交互,挖掘读者动态阅读兴趣,分析读者借阅偏好,从而提供更加精准的图书推荐服务.对武汉轻工大学图书馆5年真实数据的实验结果表明,该文所提出的DRPM模型比基线模型在命中率(HR)和归一化折损累积增益(NDCG)评价指标上有较大提升.
Traditional recommendation algorithms face problems such as sparse data and weak generalization ability,resulting in low recommendation accuracy.In the past decade,deep learning has become the most popular research direction due to its strong expression ability and flexible model structure.In this paper,a deep reader preference ming for book recommendation(DRPM)is proposed.Natural language processing technology is used to extract rich semantic features of books.The LSTM is used to model readers historical borrowing records and analyzes readers interests.In addition,attention mechanism is introduced to realize the effective interaction between books and readers,excavate readers dynamic reading interests and analyze readers borrowing preferences,so as to provide more accurate book recommendation services.Result of experiments on dataset of Wuhan Polytechnic University Library shows that the DRPM model proposed in this paper has a greater improvement in the evaluation of HR and NDCG indicators compared to the baseline model.
作者
刘园园
LIU Yuanyuan(Library,Wuhan Polytechnic University,Wuhan 430023,China)
出处
《西南大学学报（自然科学版）》
CAS
CSCD
北大核心
2023年第4期201-209,共9页
Journal of Southwest University(Natural Science Edition)
基金
国家自然科学基金项目(61906140)
湖北省自然科学基金杰出青年项目(2020CFA063)
武汉轻工大学高等教育研究一般项目(2020GJKT013).
关键词
推荐算法
深度学习
自然语言处理
长短时记忆网络
注意力机制
recommendation algorithm
deep learning
natural language processing
Long Short-Term Memory
attention
