基于ARIA的K均值聚类算法研究
|
摘要
针对传统K均值聚类算法对初始聚类中心敏感,易陷入局部最优和对大数据集聚类速度慢的缺点,将ARIA与Kmeans算法相结合,提出了一种ARIA-Kmeans算法,即基于自适应半径免疫的K均值聚类算法。首先利用自适应半径免疫算法对数据进行预处理,产生能够代表原始数据分布以及密度信息的内部镜像数据;然后用K均值聚类算法对其进行多次聚类,获得最佳聚类中心,并将其作为初始聚类中心,推广到全部数据优化聚类效果;最后对其结果进行评价。实验结果表明,相对于传统Kmeans算法,新算法在保证聚类准确度的前提下,提高了算法运行的时间效率和稳定性。
Considering the shortcomings of traditional Kmeans algorithm,which is sensitive to initial clustering center and easy to fall into local optimization,an ARIA-Kmeans algorithm is proposed by an idea that combines adaptive radius algorithm( ARIA) with Kmeans clustering algorithm,which is called Kmeans clustering algorithm based on ARIA. Firstly,the adaptive radius immune algorithm is used to preprocess the data to generate internal images data that can represent the original data distribution and density information. Then,the Kmeans clustering algorithm is used to cluster the internal images data several times,and the obtained best center is taken as the initial cluster center,which is extended to all data to obtain the global optimal results. Finally,the results are estimated by corresponding indexes. Experimental results show that the new algorithm achieves better results than the traditional Kmeans algorithm in terms of both efficiency and stability,while ensuring the accuracy of clustering.
Considering the shortcomings of traditional Kmeans algorithm,which is sensitive to initial clustering center and easy to fall into local optimization,an ARIA-Kmeans algorithm is proposed by an idea that combines adaptive radius algorithm( ARIA) with Kmeans clustering algorithm,which is called Kmeans clustering algorithm based on ARIA. Firstly,the adaptive radius immune algorithm is used to preprocess the data to generate internal images data that can represent the original data distribution and density information. Then,the Kmeans clustering algorithm is used to cluster the internal images data several times,and the obtained best center is taken as the initial cluster center,which is extended to all data to obtain the global optimal results. Finally,the results are estimated by corresponding indexes. Experimental results show that the new algorithm achieves better results than the traditional Kmeans algorithm in terms of both efficiency and stability,while ensuring the accuracy of clustering.
引文
[1]曹跃,王雅琳,何海明,等.Canopy-Kmeans聚类和组合优化的铁矿预配料智能调度[J].控制理论与应用,2017,34(7):947-955.
[2]刘倩颖,阮应君,时翔,等.基于kmeans聚类与BP神经网络算法的办公建筑逐时电负荷预测[J].热能动力工程,2018,33(3):138-144.
[3]王亚涛,王新珩,董育宁,等.基于Kmeans和动态WKNN的两层Wi-Fi改进定位方法[J].南京邮电大学学报:自然科学版,2017,37(5):41-47.
[4]洪月华.基于MPI蜂群K均值聚类算法并行化计算[J].计算机工程与设计,2017,38(12):3339-3343.
[5]廖伍代,朱范炳,王海泉,等.基于人工蜂群优化的K均值聚类算法[J].计算机测量与控制,2018,26(4):136-138.
[6]CHEN Z,TANG T H.Coding technology in Galileo system based on chaotic spreading spectrum CDMA[J].Information&Electric Engineering,2010,8(2):25-31.
[7]LEE J H,SHIM D S.Fast acquisition of GPS L5 PRNand NH code using L1 signal for software receivers[J].International Journal of Control Automation&System,2016,14(4):1-7.
[8]DIESPOSTI R S.Global Position System(GPS)user receiver and geometric surface processing for all-in-view coherent GPS signal PRN codes acquisition and navigation solution:US,7688261B2[P].2010-03-30.
[9]BEZERRA G B,BARRA T V,CASTRO L N D,et al.Adaptive radius immune algorithm for data clustering[C]//Proceeding of 4th International Conference on Artificial Immune System(ICARIS 2005),Banff,Alberta,Canada,August 14-17,2005:290-303.
[10]JERNE N K.Towards a network theory of the immune network for data analysis[J].Data mining:a heuristic approach,2001,125C(2):373-383.
[11]CASTRO L N D,ZUBEN F J V.aiNet:An artificial immune network for data analysis[J].Data Mining A Heuristic Approach,2002,40(11):1641-1645.
[12]FRANCA F O D,COELHO G P,CASTRO P A D.Conceptual and practical aspects of the aiNet family of algorithms[J].International Journal of Natural Computing Research,2010,1(1):1-35.
[13]HAN J W,KAMBER M,PEI J.数据挖掘概念与技术[M].3版.范明,孟小峰,译.北京:机械工业出版社,2012.
[2]刘倩颖,阮应君,时翔,等.基于kmeans聚类与BP神经网络算法的办公建筑逐时电负荷预测[J].热能动力工程,2018,33(3):138-144.
[3]王亚涛,王新珩,董育宁,等.基于Kmeans和动态WKNN的两层Wi-Fi改进定位方法[J].南京邮电大学学报:自然科学版,2017,37(5):41-47.
[4]洪月华.基于MPI蜂群K均值聚类算法并行化计算[J].计算机工程与设计,2017,38(12):3339-3343.
[5]廖伍代,朱范炳,王海泉,等.基于人工蜂群优化的K均值聚类算法[J].计算机测量与控制,2018,26(4):136-138.
[6]CHEN Z,TANG T H.Coding technology in Galileo system based on chaotic spreading spectrum CDMA[J].Information&Electric Engineering,2010,8(2):25-31.
[7]LEE J H,SHIM D S.Fast acquisition of GPS L5 PRNand NH code using L1 signal for software receivers[J].International Journal of Control Automation&System,2016,14(4):1-7.
[8]DIESPOSTI R S.Global Position System(GPS)user receiver and geometric surface processing for all-in-view coherent GPS signal PRN codes acquisition and navigation solution:US,7688261B2[P].2010-03-30.
[9]BEZERRA G B,BARRA T V,CASTRO L N D,et al.Adaptive radius immune algorithm for data clustering[C]//Proceeding of 4th International Conference on Artificial Immune System(ICARIS 2005),Banff,Alberta,Canada,August 14-17,2005:290-303.
[10]JERNE N K.Towards a network theory of the immune network for data analysis[J].Data mining:a heuristic approach,2001,125C(2):373-383.
[11]CASTRO L N D,ZUBEN F J V.aiNet:An artificial immune network for data analysis[J].Data Mining A Heuristic Approach,2002,40(11):1641-1645.
[12]FRANCA F O D,COELHO G P,CASTRO P A D.Conceptual and practical aspects of the aiNet family of algorithms[J].International Journal of Natural Computing Research,2010,1(1):1-35.
[13]HAN J W,KAMBER M,PEI J.数据挖掘概念与技术[M].3版.范明,孟小峰,译.北京:机械工业出版社,2012.