摘要
互联网时代,个性化推荐系统逐渐被应用到各个不同的领域,随之个性化推荐算法也成为目前研究的热点。然而,传统的推荐算法往往存在着冷启动、数据稀疏等问题。本文在对传统推荐算法研究的基础上,提出了一种基于相似传播和情景聚类的协同过滤推荐算法,根据计算用户间的情景相似度对用户进行聚类,然后根据相似传播原理找出目标用户更多的最近邻居,最后根据预测目标用户对项目的评分进行推荐。借助网上公共数据集在Matlab上实现了该算法并验证了算法的有效性。实验结果表明,本文所提算法的准确性相比传统算法有所提高,同时缓解了传统推荐算法存在的冷启动和数据稀疏性等问题。
In the age of the Internet era,the personalized recommendation system gradually is applied to different fields and recommendation algorithm has become a research hot spot at present. Traditional recommendation algorithm,however,often has some problems,for example a cold start,sparse data. In this paper,on the basis of researches on traditional recommendation algorithm,this paper proposed a collaborative filtering recommendation algorithm based on similarity propagation and context clustering. Computing the similarity between user for user clustering,then the paper found more nearest neighbors of target users,according to the similarity propagation to finally,it recommended projects according to the forecast target user's ratings. With the help of online public data,the paper implemented the proposed algorithm and verified the effectiveness of the proposed algorithm on Matlab. experiment showed that the accuracy of the proposed algorithm compared with the traditional algorithm was higher,and the proposed algorithm relieved the problems of traditional recommendation algorithm,such as the cold start and sparse data,etc.
出处
《现代情报》
CSSCI
北大核心
2016年第11期50-54,共5页
Journal of Modern Information
基金
江西省教育科学规划项目"面向绩效的高校管理决策行为关系研究"(项目编号:16YB022)的部分研究成果
关键词
相似传播
情景聚类
协同过滤
推荐算法
similarity propagation
context clustering
collaborative filtering
recommendation algorithm