A Clustering Algorithm Based on Information Visualization

This p aper st udies a clustering algorithm b ased on i nformation visulization. In th is alg orithm, through a nonlinear ma pping (NLP), s ome hig h-dimensional and complicated f eature data i s transformed into low-dimensi onal feature d ata, such as o ne, two and three dim ensionality. I ts mai n aim is that the geometry i mage in high-dimensional s pace is ma pped in to one, two a nd three dimensional i mage in low-dimensional s pace, and the inheren t data " structure" is approximat ely preserved after mapping. The simulated results show that the algorithm presented here is feasible and effective wi th d irect ob servation and im age et al. It describes well non linear character for h ighdimensional feature data.