Artificial intelligence

Scientific journal

ISSN 2710-1673

ONLINE: ISSN 2710-1681

Select your language


Комбинаторная оценка влияния уменьшения информационного покрытия классов на обобщающую особенность 1NN алгоритмов классификации

Kapustiy B.1, Rusin B.2
1 Національний університет “Львівська політехніка”
2 Фізико-механічний інститут ім. Г.В. Карпенка НАН України
rusyn@ipm.lviv.ua

Full text (PDF)

UDC: 004.93+519.2
Publication Language: Ukrainian
Stuc. intelekt. 2008; 13(1):49-54

Abstract: In this paper the combinatorial approach for definition of the class size reduction influence on correct recognition probability when one uses 1NN classifier. The recognition results are familiar before database class size reduction for every test object. The probability that recognition system has peculiarity to retain the recognition rate after class size reduction has been determined. The probability definition task that negative recognition results after class size reduction will become positive has also been solved.

Keywords:

References:

View full text (PDF)