An improved multi-level clustering algorithm based on k-prototype
Jing Liu · Journal of Hohai University · 2007
When k-prototype algorithm was employed to process complex data sets, some clusters with low-purity always occurred.An improved multi-level clustering algorithm based on k-prototype was proposed to tackle the above problem.In order to improve the quality of clustering,re-clustering was performed on those clusters with low-purity through automatic selection of attributes.Experimental results on data set from UCI machine learning repository show that the present algorithm can improve the clustering quality evidently,and it is also suitable for data abstraction.