Modified genetic programming combining with particle swarm optimization and performance criterion in solar cell fabrication

Abstract

This study describes the design and development of the novel model for the process optimization of solar cell fabrication. The model performance can affect the result of the physical experiment in the solar cell fabrication because the high accuracy model can provide the closer result to the output efficiency of the physical experiment. In this study, genetic programming (GP) based modeling technique was developed for the process simulation. GP is a global modeling technique, so it is suitable for process data modeling. This study describes the modified GP algorithm to solve the constant terminal problem. In the traditional GP, the constant term can be randomly selected within the fixed range when the structure is changed. Therefore, the variation ratio of the constant is too low to fit the model well. In this study, the novel GP is proposed. The method includes particle swarm optimization (PSO) to optimize the constant term in the terminals. PSO is a strong searching algorithm without a high computation cost. Actually, through the simulation results, the modeling performance and speed can be improved by the proposed GP. Because by the proposed modeling method, the structure and parameters of the model can be optimized simultaneously, the proposed method can be used as the new global modeling approach.

Genetic based feed-forward neural network training for chaff cluster detection

“…The Proposed performance criterion for model selection is based on the GMDH to minimize the error and at the same time to prevent over-fitting of the empirical data set [8]. Firstly, a given data is divided into three groups; training data, test data and unseen data.…”