Decision support for preliminary medical diagnosis integrating the data mining methods

Abstract

In this paper, we investigate the human physiological data that describe the human functional state. Some physiological features are measured by changing the physical load. Two groups are analysed: sportsmen and ischemic heart-diseased men. The multidimensional data, are formed from the parameters of the physiological features. These data are classified using support vector machine (SVM). When the data are classified, it is important to determine so called the decision surface that divides the data into two classes. It is impossible to comprehend the decision surfaces in the n-dimensional space (n > 3). So, we propose to project the multidimensional data on to a plane, and then to determine the decision surfaces in two-dimensional space. To make a preliminary diagnosis of the new patient, it suffices to map the point, corresponding to this patient, among the previously mapped points. A decision on the possible health troubles is made via a visual analysis of the position of the point in respect of the decision surfaces.