Дослідження впливу типу метрики на точність кластерізації нейронною мережею Кохонена у задачі медичного діагностування за аналізом крові

Abstract

In current article the known metrics were analyzed and the influence of the metrics type on the accuracy of the clustering performed by Kohonen neural network was investigated. It was established that by using the weighted Euclidean distance as a metric in comparison with the others, the highest diagnostic accuracy in the task of medical patients’ diagnostics by a blood test using a Kohonen neural network is achieved.

Kohonen neural network, metrics, distance measure, clustering, medical diagnostics