Integration of classification and visualization for diagnosis decisions

Abstract

A method of integration of classification and visualization for making diagnosis decisions is proposed. From the table of the numerical medical multidimensional data that characterise the human health state, it is difficult to decide which class the data belong to. Various classification methods can be used. However, classification errors are inevitable. Therefore, it is advisable to combine the data analysis methods of different types: in addidon to classification, visualisation of the multidimensional data (projection of the data onto a plane) may be performed. The visual analysis allows us to assign data points to one of the possible classes in a simple visual way. The results can help medics to make a preliminary diagnosis successfully or to have some doubts on the former diagnosis.