WEIGHTED CLUSTERING ALGORITHM FOR THE DATA WITH MIXED ATTRIBUTES

摘要：

　　Based upon W-k-means framework, a new clustering algorithm is proposed.In the algorithm, the concept of distribution centroid is introduced to represent the centre of cluster with categorical attributes, then distribution centroid and mean is combined to represent the centre of cluster of mixed numeric and categorical attribute data.A new dissimilarity measure, which takes into account the influence of different attributes in clustering process, is used to calculate the distance between data objects and the centre of cluster.In addition,the weight strategy of the W-k-means framework is used to assess the influence of different attributes.The proposed algorithm is verified via a series of experiments on real world datasets.Better clustering accuracy is demonstrated in comparison with those of traditionalclustering algorithms.

关键词： Cluster analysis Numericattribute data Categorical attribute data W-k-means algorithm

作者: Yiyang Wang Li Wang Zhong Qian Bo Xu Chao Lei Yao Zhong Yue You

作者单位: School of Economics and Management, Beihang University, Beijing 100191, China

会议类型: 国际会议

会议名称: The 12th International Conference on Industrial Management(第十二届工业管理国际会议)

会议地点: 成都

会议语种:英文

页码: 333-338

在线出版日期: 2014-09-03（万方平台首次上网日期，不代表论文的发表时间）