摘要
[目的/意义]旨在利用属性信息优化图书推荐模型,提升模型个性化推荐能力。[方法/过程]充分挖掘读者、书籍属性信息对读者偏好的影响,提出一种融入属性信息的度量分解模型,将读者、书籍置入度量空间,利用属性重构偏好向量,通过学习度量距离约束偏好向量分布以生成预测评分,实现个性化书籍推荐。在Book-Crossing数据集中进行相关实验,评分预测能力。[结果/结论]该模型预测准确度优于对比的其他基线模型,准确度提升比率为7%-15%。
[Purpose/significance]The paper intends to optimize book recommendation model through attribute information and promote individual recommendation ability of the model.[Method/process]The paper adequately exploits the impact of attribute information of readers and books on reader preferences proposes a metric factorization model integrated attribute feature puts readers and books into metric space reconstructs preference vector with attribute information constraints preference vector and obtains prediction score through learning distance measure so as to realize individual book recommendation.Finally it is put into Book-Crossing dataset to conduct an experiment so as to evaluat the predictive ability.[Result/conclusion]The results show that the prediction accuracy of the model is better than the other baseline models and the accuracy improvement rate is 7%-15%.
作者
于慧
温廷新
黄培峰
Yu Hui;Wen Tingxin;Huang Peifeng(System Engineering Institute Liaoning Technical University,Huludao Liaoning 125105;Huludao Industrial Informatization Development Center,Huludao Liaoning 125105;Department of Basic Bohai Shipbuilding Vocational College,Huludao Liaoning 125105)
出处
《情报探索》
2022年第4期114-121,共8页
Information Research
基金
辽宁省社会科学规划基金项目“辽宁新型城镇化评价指标体系研究”(项目编号:L14BTJ004)成果之一。
关键词
图书推荐
度量分解
属性信息
坐标构建
偏置约束
book recommendation
metric factorization
attribute information
coordinate construction
bias constraint
