Research On B-spline Network Model Optimized By Modified Artificial Bee Colony Algorithm

Time series forecasting is very common in daily life.Analysis and research forecasting plays a very important role in production and life.Artificial neural networks have strong nonlinear processing capabilities and provide a new way to solve nonlinear time series problems.Then,B-spline network has a simple structure and fast learning,which has good theoretical and practical value.This paper mainly studies the B-spline network model.First we proposes a modeified artificial bee colony algorithm(MABC).and then optimize the B-spline network model by MABC algorithm,and finally,the B-spline network model based on the MABC algorithm is applied to the empirical analysis of the daily closing price of Shanghai Securities Composite Index and the daily average air pressure of Hangzhou Station.The simulation results show that the B-spline network model based on the MABC algorithm in this paper has good generalization performance in different application scenarios.The main work of this paper are as follows:First,aiming at the problem that the ABC algorithm tends to converge prematurely and fall into local extreme values,the ABC algorithm is improved in two aspects:(1)optimize the local search method based on the global optimal strategy to avoid premature convergence;(2)based on the reverse roulette selection strategy to optimize the selection probability of the following process and maintain the diversity of the population.In addition,the MABC algorithm used Matlab program to conduct numerical experiment analysis through fitting test function.Second,aiming at the problem of poor prediction effect of B-spline network and easy to fall into local optimum,the B-spline network model is improved in two aspects:(1)The position of the B-spline network node is optimized by the random disturbance search strategy.The uneven placement of nodes replaces the uniform increase of nodes;(2)The expansion coefficients of the B-spline network are optimized based on the MABC algorithm,and the global optimal solution of the expansion coefficients and node vectors is obtained through iterative optimization.Third,in order to verify the effect of the B-spline network model based on the MABC algorithm,on the one hand,this paper perform numerical simulation analysis,and compares and analysis among the B-spline network model,the knot-optimizing B-spline network model,and the MABC-optimizing B-spline network model.On the other hand,these models are applied to the forecast of the daily closing price of the Shanghai Securities Composite Index in the financial sector and the daily average pressure of Hangzhou Station in the meteorological sector.The research results show that the MABC algorithm better balances the exploration and development capabilities,and improves the calculation accuracy and stability of.The node positions and weight coefficients of the B-spline network model are used as independent decision variables,and the B-spline network model further optimized by the MABC algorithm has a smaller prediction error.The application results of this model in the financial and meteorological fields show that the B-spline network prediction model optimized based on the MABC algorithm has better generalization performance and has good prediction effects in different fields.