Application of discriminant analysis and decision tree method for diagnosis of ophthalmic diseases

The article is devoted to the development of algorithms for classification and diagnosis of eye diseases. This task is important in practice. A classification algorithm obtained can be used in practice in future. Data were provided by St. Petersburg State Enterprise “Diagnostic center N 7 (eye)” for adults and children. The presented solution consists of several parts. A lot of preparatory work was done with the data (outliers were selected and excluded from the sample as well as the training set consisting of more than 700 observations was created). Several classification schemes were developed. Different statistical methods were used. A mixed method (discriminant analysis and classification trees were included) for each of the classification schemes was created. It provides the highest accuracy. The classification scheme with the best accuracy was selected. The resulting algorithm has been implemented in C++. The accuracy of the classifier was 72%. Thus an important practical problem has been solved which will help speed up and simplify the work of doctors. Bibliogr. 7. Il. 3.