Research On Mixed Model Based On Least Squares Support Vector Machine And Markov Model

With the rapid development of the Internet+ era and smart terminals.Data prediction and mixture models based on multi angle analysis are playing a more and more important role in scientific research and social practice.Artificial intelligence based on machine learning which has brought more happiness feeling in people's lives.It is the inevitable trend of the development of human society in modern times.Classification is favored by researchers as the core of machine learning.So more scholars begin to study classification in many application areas.In the research process,they introduced the intelligent classical algorithm based on support vector machine,and solved the common prediction problems in practical problems.However,the calculation of the support vector machine model is more complex and the resulting in the promotion is limited.Then the nonlinear least square support vector machine that solves the complexity problem is come into being.Markov chain is a stochastic process of discrete events,which is used to solve queuing problems and statistical problems in modeling.As a common prediction model,Markov chain plays an important role in the stochastic dynamic model,which is mainly used to analyze the transfer probability in the development process of things.These two models have been widely applied in scientific research and practice,and have been achieved remarkable results.The support vector machine and the Markov chain model have been widely used in many fields,but the traditional model is single and more suitable in single field,which is lack of innovation and extensiveness.While the hybrid model can overcome the singularity preferably,and improve the prediction accuracy in the prediction model through multi-level,wide angle and wide area operation and analysis.Therefore,the paper focuses on the hybrid model of least squares support vector machine and Markov model,so as to solve the problem of low accuracy of single forecast model in forecasting process and to realize the optimization model.The simulation experiment is carried out on the basis of the highway passenger volume in Lanzhou,Gansu province.The main research contents of this paper are as follows:(1)Reviewed the development course of least squares support vector machine,Markov chain,as well as the current research situation at home and abroad,and then expounded on the definition of VC dimension,the boundary of generalization,the estimation and selection of the model in the statistical learning theory.Meanwhile,Markov process is explained by Markov chain's positive stability and finite nature.In this paper,kernel functions are used to simplify the inner product operation,which effectively avoids the complexity of computation because of the relatively high dimension.(2)From the four angles of the data preprocessing,model implementation process,parameter selection and parameter calibration,a least square support vector machine model is establish,and Markov model based on the residual sequence is also investigated.In view of the low precision of single model,a hybrid model based on least squares support vector machine and Marko model is established in this paper.(3)Taking highway passenger traffic data in Lanzhou,Gansu province as the basic data,PSO algorithm is used to optimize the processing of relevant data,and compared simulation results from different models.The simulation experiments show that the accuracy of hybrid prediction model is better than the single ARIMA model or GM-Markov model,which provides strong practicability to solve the problem of low accuracy of the prediction data of the single model effectively.