迁移学习优化TB算法的个性化推荐模型  

Personalized Recommendation Model Based on Transfer Learning Optimized TB Algorithm

作　　者：葛建霞[1] Ge Jianxia(Mingda Polytechnic Insitiute,Yancheng Jiangsu 224300,China))

机构地区：[1]明达职业技术学院信息传媒系,江苏盐城224300

出　　处：《科技通报》2017年第5期138-141,共4页Bulletin of Science and Technology

摘　　要：针对标准Trada Boost算法在个性化推荐的应用中还存在精度不高的问题,本文提出了一种基于迁移学习优化Trada Boost算法的个性化推荐模型。首先采用迁移学习对Trada Boost算法的类簇进行平滑处理,即使用聚类算法对用户—项目评分数据进行预处理,并产生预选最近邻用户,接着使用基于内存的协同过滤算法在聚类结果数据上对评分数据进行平滑,最后采用改进的Trada Boost算法构建个性化推荐模型。通过仿真实验表明,本文提出的改进算法相比较目前常用的算法具有更高的个性化推荐精度。In order to solve the low accuracy problem of the standard TradaBoost algorithm in theapplication of personalized recommendation,this paper proposes a personalized recommendation modelbased on the transfer learning optimization TradaBoost algorithm.First,the transfer learning method isused to smooth the cluster of the TradaBoost algorithm,that is,the clustering algorithm is used topreprocessing the user-item rating data,and the nearest neighbors are pre-selected.Second,the ratingdata of the clustering results is smoothed by using the memory-based collaborative filtering algorithm.And third,a personalized recommendation model is constructed based on the improved TradaBoostalgorithm.Simulations show that the improved algorithm proposed in this paper has higherrecommendation accuracy than that of the commonly used algorithm.

关 键 词：个性化推荐 迁移学习 类簇平滑 协同过滤 改进TradaBoost算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]