Weighted Clustering and Evolutionary Analysis of Hybrid Attributes Data Streams

Abstract

It presents some definitions of projected cluster and projected cluster group on hybrid attributes after having given some definitions on ordered attributes and sorted attributes to solve clustering analysis problem of infinite hybrid attributes data streams in finite space. In order to improve the clustering quality of hybrid attributes data streams, it presents a two-step projected clustering method, which can often make better clustering effects in two simulated experiments although it is very simple. At last, it gives a dividing and merging framework of infinite hybrid attributes data streams. In order to implement this framework, it presents 8 properties in Section IV, some data structure definitions and 15 algorithms in appendix. The framework is verified and these algorithms are tested by German data set with a better clustering quality than WKMeans sometimes if having set right parameters.