Tác giả: Nghị Vĩnh Khanh, Dr. Richard H. Fowler, Dr. Wendy A. Lawrence-Fowler, Dr. Zhixiang Chen. (Người hướng dẫn khoa học).

University of Texas-Pan. Năm XB: 2013. Mô tả: 94Tr, kích thước: 30cm

Visual Data Mining have proven to be of high value in exploratory data analysis and data mining because it provides an intuitive feedback on data analysis and support decision-making activities. Several visualization techniques have been developed for cluster discovery such as Grand Tour, HD-Eye, Star Coordinates, etc. They are very useful tool which are visualized in 2D or 3D; however, they have not simple for users who are not trained.

This thesis proposes a new approach to build a 3D clustering visualization system for document clustering by using k-mean algorithm. A cluster will be represented by a neutron (centroid) and electrons (documents) which will keep a distance with neutron by force.

Our approach employs quantified domain knowledge and explorative observation as prediction to map high dimensional data onto 3D space for revealing the relationship among documents. User can perform an intuitive visual assessment of the consistency of the cluster structure.