摘要
以黑龙江省中部地区2004年的气温和降水观测数据为实例,对数据挖掘中气象数据的概化问题进行研究.调查分析了概化方法和数据特征,确定采用模糊C均值聚类算法.重点研究了算法中最佳聚类数的确定,选取多种广泛认可的指标,对聚类结果进行有效性对比分析,采用综合最优的聚类数为最佳聚类数.研究发现这种方法聚类结果合理且具有较强的可拓展性.
In this paper, meteorological data generalization is performed by using the temperature and rainfall observation data of central Heilongjiang in 2004. After a general analysis, the cluster algorithm of fuzzy C-means is used for classifying experimental data. How to determine the optimal class number is critical issue for data generalization. This paper has comparatively selected several meaningful and widely accepted indexes for validating the results calculated with different class numbers. The optimal class number can satisfy these indexes to a large extent. It turns out that this method is effective and scalable for computing the optimal class number and obtaining reasonable cluster results.
出处
《华中师范大学学报（自然科学版）》
CAS
CSCD
2008年第3期490-493,499,共5页
Journal of Central China Normal University：Natural Sciences
基金
国家重点基础研究发展规划项目(2006CB701305)
国家高技术研究发展计划项目(2008AA12Z201)
关键词
模糊聚类
气象数据
FCM
最佳聚类数
fuzzy cluster
meteorological data
FCM
optimal class number
