基于层次化上下文因式分解机的推荐系统  

Recommendation System Research Based on Hierarchical Context Factorization Machines

作　　者：秦大路 李晓宇[1] 

机构地区：[1]郑州大学信息工程学院,郑州450001

出　　处：《河南师范大学学报（自然科学版）》2015年第2期147-151,共5页Journal of Henan Normal University(Natural Science Edition)

基　　金：国家自然科学基金(61073023)

摘　　要：在基于协同过滤的推荐系统中,因式分解机模型是基于矩阵分解的一般化模型,不需要特定支持向量,可直接应用于回归和分类中,并能更准确地处理稀疏矩阵.通过对其进行改进,在不提高时间复杂度的同时考虑上下文环境,并对上下文进行层次化处理.通过两组真实数据集,在不同的指标下进行实验.最后证实改进后的模型,在准确率和学习速率上优于原有模型.In the recommender systems based on collaborative filtering,the factorization machines model（FM）is a generalized model based on matrix factorization method which needs no specific support vectors.It can be applied in regression and classification and process sparse matrix exactly.By modifying FM it＇s possible to consider the context and implement hierarchical processing the context without improving the time complexity.Experiments on two group of real data have been done under different indices.It＇s proved that the modified model is better than the previous models in accuracy rate and learning speed.

关 键 词：推荐系统 协同过滤 矩阵分解 上下文环境 因式分解机 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]