An optimised hybrid group method in data handling (GMDH) network

Abstract

A novel modular optimized hydrid Group Method in Data Handling (GMDH) network is proposed in this paper. A standard GMDH network is optimized using the Discrete Differential Evolution (DDE) algorithm for an optimized network structure, and Singular Value Decomposition (SVD) is further used for coefficient calculations of the network. The developed DE-GMDH algorithm is tested for fitness accuracy, memory usage and maximal error on a manufacturing problem.