摘要
个性化推荐使得用户从浩瀚信息检索查找中解放出来,成为一种继搜索引擎之后获取信息的重要方式。协同过滤因为其算法简单,能够处理复杂对象,并且推荐效果优异,成为个性化推荐中最成功和应用最广泛的技术。但随着推荐系统规模扩大,协同过滤受到了数据稀疏性、冷启动和可扩展性等瓶颈问题严重挑战。本文总结了传统协同过滤推荐技术流程,重点研究了解决协同过滤瓶颈问题的方案,分析了它们各自的优缺点,便于后续实现协同过滤推荐系统时方案的选择和使用。
The personalized recommendation, through which users are free from the vast information retrieval, is an important way to obtain information after search engine. Collaborative filtering, a simple algorithm, is able to handle complicate objects and have good reommendations, becoming the most successful and the most widely used technology in personalized recommendation. However, collaborative filtering suffers serious challenges from data sparseness problem, cold start problem and scalability problem with the expansion of recommendation system. This article summarizes the traditional collaborative filtering recommendaion process,mainly researches solutions of problem of collaborative filtering bottleneck and analyzes the advantages and disadvantages of the solutions, In order to facilitate the follow-up selection of the solutions.
出处
《软件》
2012年第12期315-321,共7页
Software
关键词
个性化推荐
协同过滤
数据稀疏性问题
冷启动问题
可扩展性问题
personal recommendation
collaborative filtering
data sparseness problem
cold start problem
scalability problem
