A New Initialization Method for Clustering Categorical Data

Shu Wu, Qingshan Jiang, Joshua Zhexue Huang

Abstract: Performance of partitional clustering algorithms which converges to numerous local minima highly depends on initial cluster centers. This paper presents an initialization method which can be implemented to partitional clustering algorithms for categorical data sets with minimizing the numerical objectivefunction. Experimental results show that the new initialization method is more efficient and stabler than the traditional one and can be implemented to large data sets for its linear time complexity.