基于混合聚类的个性化推荐算法  

A Personal Recommendation Algorithm for Digital Library Based on Hybrid Clustering Technology

作　　者：成桂兰[1] 刘旭东[1,2] 陈德人[3] 

机构地区：[1]烟台职业学院信息工程系,山东烟台264670 [2]烟台职业学院科研处,山东烟台264670 [3]浙江大学计算机科学与技术学院,浙江杭州310027

出　　处：《武汉理工大学学报（信息与管理工程版）》2011年第3期379-381,434,共4页Journal of Wuhan University of Technology：Information & Management Engineering

基　　金：国家科技支撑计划基金资助项目(2008BAH21B03)

摘　　要：针对传统协同过滤算法存在的数据稀疏性问题,提出了一种基于聚类技术的推荐算法。该算法将SOM与K-means技术相结合对图书资源进行聚类,缩小了需要预测的图书资源数目和最近邻居的搜索范围,达到了为读者提供符合其偏好特征的图书资源的目的。实验结果表明,改进后的算法较好地解决了数据稀疏性问题,提高了推荐系统的推荐质量。Considering the problems of the sparsity of rating data in traditional collaborative filtering,a recommendation algorithm based on clustering technology was proposed.This algorithm combined the library resources by combining the SOM technology with the K-means technology.The number of library resources which need forecasting and the scope of neighbor search is reduced.A satisfactory effect on the recommendation was achieved.The experimental results indicate that the problems of sparseness are efficiently resolved,and the quality of recommendation is improved.

关 键 词：协同过滤 数字图书馆 SOM K-MEANS 数据稀疏性 平均绝对偏差 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]