Technical gazette, Vol. 23 No. 6, 2016.
Original scientific paper
https://doi.org/10.17559/TV-20160531025240
Polu-nadzirana propagacija afiniteta temeljena na vršnoj gustoći
Limin Wang
; School of Management Science and Information Engineering, Jilin University of Finance and Economics, 3699 Jingyue Street, Changchun 130117, China
Xing Tao
; School of Management Science and Information Engineering, Jilin University of Finance and Economics, 3699 Jingyue Street, Changchun 130117, China
Xuming Han
; School of Computer Science and Engineering, Changchun University of Technology, 307, Teaching Building, No. 7186, Weixing Road, Changchun 130012, China
Jialing Han
; School of Computer Science and Engineering, Changchun University of Technology, 307, Teaching Building, No. 7186, Weixing Road, Changchun 130012, China
Ying Liu
; School of Computer Science and Engineering, Changchun University of Technology, 307, Teaching Building, No. 7186, Weixing Road, Changchun 130012, China
Guangyu Mu
; School of Computer Science and Engineering, Changchun University of Technology, 307, Teaching Building, No. 7186, Weixing Road, Changchun 130012, China
Abstract
Zbog nezadovoljavajućeg učinka grupiranja (klasteriranja) pomoću algoritma grupiranja propagacijom afiniteta (AP - affinity propagation) u slučaju nizova podataka složene strukture, u radu se predlaže polu nadzirani algoritam grupiranja propagacije afiniteta temeljen na vršnoj gustoći (SAP-DP). Taj algoritam primjenjuje novi algoritam vršne gustoće (DP - density peaks) čija je prednost višestruko grupiranje uz polu-nadziranje, izgradnja udvojenih ograničenja zbog usklađivanja s matricom sličnosti, a zatim izvršenje grupiranja propagacijom afiniteta. Rezultati simulacijskih eksperimenata potvrdili su da je grupiranje predloženim algoritmom učinkovitije od grupiranja konvencionalnom propagacijom afiniteta (AP).
Keywords
polu-nadgledano učenje; propagacija afiniteta; vršna gustoća; udvojena ograničenja
Hrčak ID:
169536
URI
Publication date:
29.11.2016.
Visits: 2.273 *