改进型特征权重自调节K-均值聚类算法  

K-means clustering algorithm with an improved feature weight self-adjustment mechanism

作　　者：支晓斌[1] 许朝晖[2] 

机构地区：[1]西安邮电大学理学院,陕西西安710121 [2]西安邮电大学通信与信息工程学院,陕西西安710121

出　　处：《西安邮电大学学报》2014年第6期26-31,共6页Journal of Xi’an University of Posts and Telecommunications

基　　金：陕西省自然科学基金资助项目(2014JM8307)

摘　　要：针对特征权重自调节K-均值聚类(FWSA-KM)算法对噪声敏感的问题,提出一种改进型特征权重自调节K-均值聚类(IFWSA-KM)算法。用一种非欧氏距离代替FWSA-KM算法中的欧氏距离,以增加聚类算法的抗噪声性能。通过用人工数据和真实数据的对比性实验,可验证IFWSA-KM算法的有效性。K-means with a feature weight self-adjustment mechanism（FWSA-KM）clustering algorithm is sensitive to noise.Therefore K-means with an improved feature weight selfadjustment mechanism（IFWSA-KM）clustering algorithm is proposed in this paper.IFWSA-KM clustering algorithm can have some anti-noise performance by using a non-Euclidean distance.The effectiveness of IFWSA-KM algorithm is demonstrated by comparative experiments on synthetic and real data.

关 键 词：聚类算法 特征权重 鲁棒性 非欧氏距离 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]