Research And Implementation Of Short-term Traffic Flow Forecasting Based On Support Vector Machine Regression

Short term traffic flow forecasting techniques are important research of intelligenttraffic control and vehicle-induced field, the use of the actual project, it is very important asthe field of intelligent transportation basic theory, analysis and forecasting of traffic flow,can help cities Intelligent Transportation induced, enabling users to select the optimal path.Support vector machine learning and less learning problems in the face of thetraditional face of difficulties in the local minima problems, small sample study of machinelearning methods can be well resolved; can better solve the short term traffic flowregression to predict the subject.Short-term traffic flow forecasting based on support vector machine regression,in-depth discussion and research on short-term traffic flow forecasting methods and theorieson the basis of the analysis of traffic flow data, and based on actual data and verify theproposed prediction model based on an improved support vector machine regression withthe actual feasibility of the application.The main work is as follows:(1) Digital signal wavelet denoising owned by one of a series of pre-processingoperations, reduce noise and to lay the foundation for the establishment of the traffic flowforecasting model;(2) According to the basic principles of support vector machines, based on supportvector machine regression of short-term traffic flow forecasting model, experimental resultsshow that support vector machine regression model is a viable, effective traffic flowforecasting model.(3) Support vector machine parameter selection optimization model. The penalty factor,kernel function parameter optimization of support vector selection plays an important rolein learning precision and generalization ability of the regression model is good or bad. Thisarticle uses the traditional particle swarm optimization algorithm, support vector regressionparameters, parameter optimization of the parameters; combined with the genetic algorithm,using an improved particle swarm optimization, simulation, improved particle swarm optimization to achieve the prediction model adaptive capacity, and improve short-termtraffic flow prediction accuracy.