Double k-nearest Neighbors of Heterogeneous Data Stream Clustering Algorithm
Huang De-ca · 2013
On the one hand,most of the existing data stream clustering algorithm can handle data with numerical attribute,but can not cope with the data containing both numeric and classification attributes.On the other hand,there is also a lot of room for heterogeneous data stream algorithms to improve standardization and clustering of data.So,double k-nearest neighbors of heterogeneous data stream clustering algorithm was proposed.The algorithm uses CluStream's online and offline framework with proposing three steps of clustering thought.Firstly,the algorithm uses double k-nearest neighbors and improved dimension distance to form micro clusters.Secondly,the algorithm uses dynamic standardization data method and cosine model based on mean value to form initial macro clusters.Thirdly,the algorithm uses cosine model based on mean value and priori clusters to do macro clustering optimization.Experimental results demonstrate that the proposed method improves clustering's accuracy and scalability.