Information Technologies and Computer Engineering

In current article a review of known metrics was held and the accuracy of the system of medical diagnostics by a blood test based on Kohonen neural network using different metrics was experimentally investigated. In this task the metric is used to determine the distance between the input vector set of general blood test indicators and a vector of cluster center values, which corresponds to a certain diagnosis of the patient. It was established, that the type of metric affects the accuracy of clustering. For the proposed system of medical diagnosis on the basis of experimental studies it was established, that the highest accuracy of diagnostics is provided using the weighted Euclidean distance

References