基于时间兴趣因子的网络表示学习图书推荐模型研究

Network Representation Learning Book Recommendation Model Research Based on the Time Interest Factor

[目的/意义]通过分析图书馆的图书流通数据,本文提出一种基于时间兴趣因子融合网络学习的图书推荐模型--TIF_N2V_CF。[方法/过程]评估用户借阅图书的时间间隔并定义兴趣因子权重,根据流通数据构建同质关系网络;网络表示学习将得到的特征矩阵输入融合推荐模型并得到推荐结果。[结果/结论]实验表明,TIF_N2V_CF模型的召回率在top z=10和z=20时分别为0.1302、0.2031,高于未引入时间兴趣因子的N2V_CF模型。TIF_N2V_CF模型将时间兴趣因子引入到网络表示学习,对融合用户和图书的特征矩阵进行相似度计算,解决图书借阅流通数据中同一时间包含多本图书借阅记录造成的难以序列化的问题,缓解数据稀疏和冷启动对模型性能的影响,提高了推荐精度。

[Purpose/significance]By analyzing the circulation data of library book borrowing,this paper proposes a library recommendation model based on time interest factor and network learning-TIF_N2V_CF.[Method/process]Evaluate the time between users to borrow books and define the weights of interest factors,constructing a homogenous relationship network based on circulation data;The network representation learning inputs the obtained feature matrix into the fusion recommendation module and obtains the recommendation results.[Result/conclusion]Experiments show that the Recall of the TIF_N2V_CF model is 0.1302 and 0.2031 at top z=10 and z=20,respectively,which is higher than the N2V_CF model without time interest factor.The TIF_N2V_CF model introduces the time interest factor into the network representation learning,and calculates the similarity of the feature matrix that combines the user and the book,the problem of difficult serialization caused by the book borrowing circulation data containing multiple book borrowing records at the same time is solved,and the impact of data sparseness and cold start on the model performance is alleviated,and the recommendation accuracy is improved.