Optimization Of BP Neural Network Based On Bat Firefly Hybrid Algorithm

With the rise of machine learning,neural network technology has once again raised attention,and has been widely used in many subject fileds,to solve problems for people.Among those subject fields,the use of neural networks for prediction is one of the most important research directions.However,in order to apply the neural network to the prediction model,the following problems must be solved:(1)Slow learning;(2)Falling into local optimum;(3)Lack of theoretical reference on the number of hidden layers and the number of neurons in each layer.This thesis proposes a new swarm intelligent optimization algorithm called hybrid bat and firefly algorithm to improve neural network.The bat algorithm is used to assist the firefly to carefully search around the global optimal individual and strengthen communication between them in order to solve the problem of slow convergence and premature under the use of firefly algorithm.Finally,Gaussian disturbance is conducted to increase the diversity of the population.It helps to improve the defects of initial solution distribution,slow convergence,easy stagnation,premature,and low resolution.Therefore,this thesis uses this algorithm to optimize BP neural network.Before training of the neural network,the initial weight and threshold of the network are trained by this algorithm,in order to obtain better initial weights and thresholds and improve network performance.The experimental results show that the neural network model has higher degree of accuracy after using the proposed algorithm,which proves the research methods and optimization measures adopted in this thesis are effective.