Grey Neural Network Model Based on Modified Particle Swarm Optimization Algorithm and Its Application

Abstract:

A grey neural network model is established with a modified particle swarm optimization (PSO) instead of the gradient correction method. The initial positions of the particles are chosen randomly according to the parameters of grey neural networks which are processed through PSO and the best individual in particle swarm algorithm is searched to improve robustness and precision of the forecasting model. Through testing the effect of solving short term order problem, the model proves to be simple with better forecast precision and of a higher approximation capability compared with back propagation(BP) neural network, grey neural network, the traditional particles warm optimizer and BP neural network. The paper presents a new method for optimizing network parameters and some new ideas for researches on forecasting model.