协同过滤技术研究综述  

Overview of Collaborative Filtering Techniques

作　　者：姚曜[1,4] 赵洪利 杨海涛[3] 张帆[4] 

机构地区：[1]装备指挥技术学院研究生管理大队,北京101416 [2]总装备部司令部,北京100720 [3]装备指挥技术学院重点实验室,北京101416 [4]92857部队

出　　处：《装备指挥技术学院学报》2011年第5期81-88,共8页Journal of the Academy of Equipment Command & Technology

基　　金：部委级资助项目

摘　　要：协同过滤作为构建个性化推荐系统的主要技术,旨在解决各领域的信息过剩问题,其主要类型包括基于记忆、基于模型以及组合协同过滤算法。分析了基于记忆的协同过滤方法的基本原理,介绍了4种典型的基于模型的方法及3类组合协同过滤算法,并从行为数据的构造、数据分割的方法及常用评估指标3方面阐述了协同过滤的算法性能的验证方法,总结了协同过滤算法在构建推荐系统时面临的问题和各类协同过滤算法的性能。单一协同过滤算法的推荐性能各有优劣,构建个性化推荐系统的关键在于：针对推荐任务的需求及数据集的特点,选择合理的推荐算法与评价指标。Aiming at solving the problem of information overload in many areas,collaborative filtering（CF）,as one of the most important techniques for building personalized recommender system,has three main categories： memory-based,model-based,and hybrid CF algorithms.In this paper,the basic procedure of memory-based CF is analyzed,four classic model-based CF algorithms and three categories of hybrid CF are introduced,and experimentation of CF algorithm is studied in three aspects： construction of action dataset,data partition,and evaluation metrics,then the problems we will meet when using CF for constructing recommender system and the performance of several classic CF algorithms are summarized.The recommender performance of each CF algorithm may be right and wrong,but the key of building a recommender system is choosing reasonable algorithm and evaluation metrics meeting the demands of recommender task and the characteristics of dataset.

关 键 词：推荐系统 个性化推荐 协同过滤 

分 类 号：TP274.2[自动化与计算机技术—检测技术与自动化装置]