摘要
聚类算法是数据挖掘中的重要技术,用于发现数据中未知的分类,是电子商务领域实现个性化推荐的重要方法。个性化推荐服务,是一种针对不同特征的用户推出适合于用户的信息及服务。文章介绍了一种K-means的聚类算法,该算法主要应用于数据挖掘领域,也在推荐系统的研究中占有重要的地位。文章还详细的分析了该算法目前存在的不足,针对该算法对初始值有一定要求这一不足,提出了一种通过预处理初始聚类中心的方法,以此来改进该算法,从实验结果可知,该改进算法能够减少迭代次数,提高计算效率。
Clustering algorithm is an important data mining technique to discover the unknown classification in data, it is also an important way to achieve personalized recommendation in e-commerce. Personalized recommendation service is a kind of method which is based on users with different features to offer the users suitable information and services.This article describes the K-means clustering algorithm, which is an important algorithm in data mining and recommendation system.This paper also analyzes the shortcomings of the algorithm. In order to address the algorithm＇s defect of the initial value sensitivity, we propose an improved algorithm of preprocessing the initial cluster centers. From the experimental results we can see that this improved algorithm can reduce the number of iterations and improve the computational efficiency.
出处
《信息通信》
2016年第1期19-21,共3页
Information & Communications
基金
上海市科学技术委员会科研项目<基于个性化推荐技术的航空移动社区服务模式研究与应用>(项目号:14DZ1101400)
同济大学研究生国际交流基金资助项目(项目批准号:201502008)的经费支持
关键词
聚类分析
数据分析
推荐系统
聚类中心初始化
迭代优化
clustering analysis
data analysis
recommendation system
initialization of cluster centers
iterative optimization
