Моделі прийняття рішень сортувальних ліній пошти на основі штучних нейронних мереж

Abstract

This report examines the results of a study of five types of artificial neural networks (ANNs) – FFNN, GRNN, PNN, RBF and LVQ – for implementing a decision-making model for automated sorting lines (ASLs) at parcel sorting centers. The aim of the study was to compare different types of ANNs and training algorithms for implementing the ASL decision-making model. Analysis of the results showed that FFNN is the most promising, as it provides the required relative probability of correct parcel classification (97 %), and its hidden (first) layer is implemented using the minimum number of neurons (28).