Presenting an improved FRBCS to handle hybrid inputs and missing features

Abstract

Evaluation of emergency medical services (EMS) is a very risky and crucial task because taking a wrong decision on that stressful circumstances leading to an uncompensated event. Recently, emergent specialists come to this conclusion that their traditional decision making process, in the situation of incomplete information, is not as accurate as they expected. This paper is aimed at developing intelligent software to take a precise decision, even in the situation of facing with uncertainty and incomplete data. In this way, a fuzzy rule-based classifier system (FRBCS) capable of working with various types of input (hybrid inputs), even in the situation of facing with missing features, is proposed and compared to the conventional machine learning algorithms. In order to exhibit effectiveness of the proposed scheme, a data set is collected with the help of “Management and Medical Emergency Center of Fars Province” containing 227 subjects with heart attack. Recorded attributes for each subject includes 33 boolean, 4 real value, and 2 nominal features (totally 39 hybrid attributes) belong to 16 essential prehospital care order. Experimental results show the superiority of the introduced scheme to the state-of-art machine learning schemes in terms of accuracy, sensitivity, and specificity.