The Study Of Support Vector Machine Time Series Prediction That Optimized By Particle Swarm Optimization

In recent years,with the continuous progress of society and the continuous increase of population,the demand for water resources is also increasing.China's water resources are too limited and unevenly distributed to be exploited and used endlessly.Therefore,it is necessary to make rational allocation of water resources.It can prevent geological disasters and water shortage caused by over-exploitation.For managers of hydrological stations,accurate observation of water level and variation trend of future water level that accurate predict by some algorithm can provide more effective scientific basis for water resources management,which is of great practical significance for rational utilization and sustainable development of water resources.On the basis of theoretical knowledge of classical particle Swarm Optimization algorithm and support vector machine(SVM),this paper proposed the use of inertia weight particle swarm optimization algorithm(Weighted Partical Swarm Optimization,W-PSO),it can optimize the two parameters of support vector machines(C) and radial basis kernel function parameters(?),on the basis of the optimal parameter combination,this paper set up the W-PSO Optimization of support vector machine(SVM)time series forecasting model,and verified by experiments.The main work of the paper is as follows:Firstly,this paper studied the basic theory of support vector machine and particle swarm optimization algorithm and analyzed the influence of punishment parameters and radial basis kernel function parameters on the accuracy and capability of the support vector machine(SVM),and based on this,come up with W-PSO algorithm for support vector machine(SVM)parameters optimization,in order to get the best parameters combination environment.Use the parameters mentioned above,this paper established a W-PSO optimized support vector machine time series prediction model and then improved the accuracy and generalization of the model predictions.Secondly,this paper used Web spider method and the Python language to extract the dongting lake water level data of the chenglingji hydrological stations as the experimental data,the experimental data respectively into the training sample set and test sample set,and the training sample set is trained with above model,then the data in the test sample set isinput into the trained prediction model to verify the feasibility of the prediction model.The experimental results show that the optimized SVM time series prediction model is superior to the traditional SVM prediction model,and this model has the advantages of high prediction accuracy and model stability.Finally,in order to truly and comprehensively evaluate the universal applicability of the prediction model,this paper used the water level data that from the dongting lake shigui mountain and dongting lake nanzhi and changjiang jianli and changjiang shashi four hydrological observation station as experimental data to verify the prediction model.It verifies the reliability and robustness of the prediction model proposed in this paper,and provided a reference basis for the further study of the water level prediction model,which has important practical significance.