Development of a Hybrid Group Method in Data Handling (GMDH) Network in C++ Framework

Abstract

This thesis describes the implementation of Group Method in Data Handling, where network structure is obtained from a data structure, that can be created with only having the basic knowable of the problem. This makes it suitable to be optimized by Evolutionary Algorithms. This in turn creates two almost independent algorithms, which gives rise to new possibilities, when it comes to optimization and parallelization. A working application that contains cross- platform optimized and parallelized implementation of Group Method in Data Handling with Discrete Differential Evolution in C++ is the main aim of this thesis.