Prediction method for optimizing gray neural network by snap-drift cuckoo search algorithm

Abstract

The invention discloses a prediction method for optimizing a gray neural network through a snap-drift cuckoo search algorithm. The method comprises the steps: acquiring a primary selection set; calculating the fitness value of each parasitic nest, and determining the minimum value of the fitness value; judging whether the updating frequency is smaller than the iteration frequency or not; if the number of iterations is smaller than the number of iterations, selecting different updating operators through a variable J to update and solve the bird eggs in each parasitic nest at present; locally searching and updating parasitic nests; selecting the worst parasitic nest based on the probability Pa; carrying out global search, and updating each abandoned parasitic nest; adjusting a performance index Pm according to the number Se of the updated solutions; judging to calculate the value of pa by using an snap mode or a drift mode according to the Pm; judging whether the updating times are smaller than the iteration times or not, if yes, returning to continue to search for the optimum, and otherwise, taking the updated parasitic nest with the minimum fitness value as the initial connection weight and threshold of the gray neural network for training until the minimum prediction precision is met.

Description

technical field [0001] The invention belongs to the fields of industry, agriculture, society, economy and other sciences, successfully solves a large number of practical problems in production, life and scientific research, and specifically relates to a prediction method of snap-drift cuckoo search algorithm optimization gray neural network Background technique [0002] Since J.M.BATES and C.WJ.GRANGER first proposed the concept of combined forecasting in 1969, combined forecasting has been favored by researchers due to its high forecasting accuracy, stability and reliability. Some results have been obtained, and many combination models have emerged, such as the combination of cluster analysis and neural network, fuzzy neural network, genetic neural network, chaotic neural network, gray neural network, and gray system and autoregressive moving average model ( ARMA) combination, combination of gray system and genetic algorithm, gray Markov process, neural network integration,...

AI Technical Summary

Problems solved by technology

[0003] However, due to the random initialization of the weights and thresholds of the gray neural network (GNN), the network is prone to fall into a local optimum, and the prediction results are different each time, and the deviation is large

Images