Linguistic Data Classification with Combined Comparison Measures

Abstract

Medical data is often imprecise, due to many reasons that can be technical or human originated. In this article, we will present a classification example where data in hand is given imprecisely. Data set presents a choice situation where medical doctor has to be able to make a decision where patient is to be sent after the surgery. Data is given linguistically, which might give the idea to use some kind of fuzzy numbers in order to decode linguistic variable into the classifiable form. In fact, this approach makes the data more imprecise and therefore harder to classify. On another hand finding of parameter values by the use of commonly used differential evolution (DE) is very time consuming. In this article, we use simple, yet effective method for decoding of linguistic data. After this we use randomly selected weights and t-norm based combined comparison measures with similarity classifier to classify data given to the correct classes. Results are compared to the existing results and method presented in this paper provides best total rate of true positive classification result of 88.89% using combination of Yager t-norm and t-conorm, whereas second highest reported best total rate of true positive classification result was 77.27% using similarity measure called Shweizer & Sklar -Łukasiewicz and Differential Evolution (DE).