Identification of Relevant Medical Parameter Values in Information Systems using Fuzzy Approach

Feature selection is an important process in machine learning. It aims to identify relevant features that describe the problem in the context of the categories distinguished in the problem domain. This paper presents the results of a study on the feasibility of using a discretization and value fuzzification process for numerical attributes to identify significant ranges of attribute values. In medicine it is applicable to identify significant values of parameters, medical indicators so that the size of medical examinations can be reduced and appropriate interpretation of symptoms can be created. The results obtained show that this is possible without losing the quality of classification while reducing the number of attributes. Moreover, by analyzing the obtained value intervals, it is possible to detail the feature selection process and detect irrelevant features not previously identified. (C) 2021 The Authors. Published by Elsevier B.V.