改进的有序聚类分析法提取时间序列转折点  

Improvement of Sequential Cluster Analysis Method for Extracting Turning Point of Time Series

作　　者：陈远中[1] 陆宝宏[2,3] 张育德[2] 周笑笑[2] 

机构地区：[1]深圳市水务工程建设管理中心,广东深圳518048 [2]河海大学水文水资源学院,江苏南京210098 [3]水文水资源及水利工程国家重点实验室,江苏南京210098

出　　处：《水文》2011年第1期41-44,共4页Journal of China Hydrology

基　　金：国家自然科学基金项目(NSFC50379008;NSFC50979023)

摘　　要：对有序聚类分析法进行改进,使其更加适用于序列转折点或突变点的提取。分别将传统和改进的有序聚类方法应用在长江下游区域年平均气温系列的转折点提取中,对两种方法提取的结果与滑动平均的结果进行比较,发现改进的方法更接近于实际。对提取点分别采用秩和检验法、游程检验法进行检验,均通过了α=0.05的置信度检验,改进后的置信度比改进前更高。Because the sequential cluster analysis method can not deal with turning point with change trend. In order to overcome the shortcomings and obtain more accurate turning points for a time series with changed trend, the conventional sequential cluster analysis approach was improved. The modified method was used to extract the turning point of annum average temperature time series of a sub-basin in Yangtze River Basin. By comparing the performance between the modified and conventional approaches, it can be found that the improved method is more close to practice. Furthermore, the obtained tuning points from the two method all passed the confidence test using rank tests and run-length tests withα=0.05, but confidence degree from the improved method is higher than the conventional approach.

关 键 词：有序类聚 转折点 时间序列 显著性检验 

分 类 号：P333.9[天文地球—水文科学]