FEATURE CONSTRUCTION WITH FUZZY DERIVATION IN DIAGNOSTICS OF UROLITHIASIS
Supervized learning as method of urolithiasis diagnostics is applied. Construction of new attributes utilizing fuzzy derivation is considered. Fuzzy derivation is applied to ordinal and real data. Cross-validation accuracy analysis is performed before and after addition of three fuzzy-based attributes. As a result of feature construction, accuracy of the decision tree classifier increased by 5 %, from 79 to 84 %. The maximum increase of accuracy was achieved on ordinal attributes.
Authors: N. I. Omirova, A. V. Tishkov
Direction: Bioengineeringl Systems in Medicine and Ecology
Keywords: Problem of classification, decision trees, fuzzy logic
View full article