基于我国居民消费价格指数的聚类分析
|
摘要
传统的k-means算法聚类由于k值的给定有一定的主观性,从而造成聚类精度的下降.为更好地利用k-means均值进行聚类,这里进行了两个方面的优化处理,分别是优化k值和改进初始聚类中心.本文以2016年我国31个省市自治区的CPI价格指数为研究对象,利用谱系图确定聚类的个数,并构建评价函数改进初始聚类中心,聚类结果显示把31个地区分为四类比较合适,并且进一步分析发现同时优化k值和优选初始聚类中心聚类结果的类间占比率较高.最后结合我国当前的发展形势对不同类别地区提出相应的价格控制建议.
The traditional k-means clustering algorithm has a certain subjectivity due to the given value of K,which leads to the decline of clustering accuracy.In order to make better use of k-means for clustering,two aspects of optimization are carried out here,namely,optimizing the K value and improving the initial clustering center.This paper takes CPI price index of 31 provinces,cities and autonomous regions in China in 2016 as the research object,uses pedigree chart to determine the number of clusters,and constructs evaluation function to improve initial cluster centers.The cluster results show that it is more appropriate to divide 31 regions into four categories,and further analysis finds that the ratio of clusters that optimize K value and initial cluster center cluster results at the same time is higher.Finally,according to the current development situation of our country,the paper puts forward corresponding price control suggestions for different types of regions.
The traditional k-means clustering algorithm has a certain subjectivity due to the given value of K,which leads to the decline of clustering accuracy.In order to make better use of k-means for clustering,two aspects of optimization are carried out here,namely,optimizing the K value and improving the initial clustering center.This paper takes CPI price index of 31 provinces,cities and autonomous regions in China in 2016 as the research object,uses pedigree chart to determine the number of clusters,and constructs evaluation function to improve initial cluster centers.The cluster results show that it is more appropriate to divide 31 regions into four categories,and further analysis finds that the ratio of clusters that optimize K value and initial cluster center cluster results at the same time is higher.Finally,according to the current development situation of our country,the paper puts forward corresponding price control suggestions for different types of regions.
引文
[1]袁利永,王基一.一种改进的半监督k-means聚类算法[J].计算机工程与科学,2011,33(6):138-143.
[2]赵姗姗.基于SPSS中系统聚类的CPI分析[D].石家庄:河南师范大学,2013.
[3]刘玉梅.基于统计分析的经济预警模型[D].重庆:重庆大学,2013.
[4]石洪景.基于因子和聚类分析的居民消费价格变动水平评价[J].西南农业大学学报(社会科学版),2012,10(10):23-27.
[5]赵亚茹.我国经济发达地区农村居民家庭消费结构影响因素分析[D].杭州:浙江工商大学,2012.
[6]詹鹏,王庚,鲍学云.我国主要城市房价的面板数据聚类分析[J].南京财经大学学报,2012,(03):14-22.
[7]胡冰川.消费价格指数、农产品价格与货币政策——基于2001~2009年的经验数据[J].中国农村经济,2010,(12):37-45.
[8]吴锐.中国消费结构差异比较实证研究[D].上海:上海交通大学,2010.
[9]谢敏.我国CPI的国际可比性研究[D].长沙:湖南大学,2009.
[10]陈琳,杨林泉.城乡居民消费结构差异的聚类分析和物价指数的影响分析[J].数量经济技术经济研究,1996,(05):73-80.
[11]陈晨.北京市居民消费价格指数波动规律及其驱动因素研究[D].北京:首都经济贸易大学,2014.
[2]赵姗姗.基于SPSS中系统聚类的CPI分析[D].石家庄:河南师范大学,2013.
[3]刘玉梅.基于统计分析的经济预警模型[D].重庆:重庆大学,2013.
[4]石洪景.基于因子和聚类分析的居民消费价格变动水平评价[J].西南农业大学学报(社会科学版),2012,10(10):23-27.
[5]赵亚茹.我国经济发达地区农村居民家庭消费结构影响因素分析[D].杭州:浙江工商大学,2012.
[6]詹鹏,王庚,鲍学云.我国主要城市房价的面板数据聚类分析[J].南京财经大学学报,2012,(03):14-22.
[7]胡冰川.消费价格指数、农产品价格与货币政策——基于2001~2009年的经验数据[J].中国农村经济,2010,(12):37-45.
[8]吴锐.中国消费结构差异比较实证研究[D].上海:上海交通大学,2010.
[9]谢敏.我国CPI的国际可比性研究[D].长沙:湖南大学,2009.
[10]陈琳,杨林泉.城乡居民消费结构差异的聚类分析和物价指数的影响分析[J].数量经济技术经济研究,1996,(05):73-80.
[11]陈晨.北京市居民消费价格指数波动规律及其驱动因素研究[D].北京:首都经济贸易大学,2014.