Study On Short-term Traffic Flow Prediction Of Urban Road

With the rapid development of China’s urbanization,urban population and motor vehicle population have increased,bringing a series of traffic problems such as heavy traffic in the city.In recent years,the extensive application of information technology in various fields has brought tremendous convenience,while the process of informatization and intelligence in the transportation field is slow.The construction of traffic network of intelligent city is urgent.At present,in view of the fact that a comparatively complete urban road traffic data gathering system has been installed in most large cities of China,it is urgent to establish scientific and effective traffic control and traffic guidance which must first be able to accurately and effectively predict shortterm traffic flow.This thesis analyzes the research status of short-term traffic flow and reveals that there are some prevailing problems such as incomplete data processing,imperfect data repair methods,and insufficient prediction accuracy.Therefore,based on the analysis of the characteristics of traffic flow,this thesis studies the basic parameters of traffic flow and the characteristics of short-term traffic flow of urban roads and possible influencing factors.Then statistical correlation analysis theory has been used to separately analyze the time and space correlation of short-term traffic flow of urban roads.Then a whole set of data processing procedures and methods is proposed to preprocess the data.In the data preprocessing,different collection methods of traffic flow data are introduced.Then,the erroneous data in the original data is deleted,the missing data is marked and the final data is repaired.The repaired data is integrated into the format required by the prediction model at a time granularity of five minutes,and the noise reduction process is performed by using the wavelet noise reduction method,which is convenient for the use in verification of subsequent examples.In the construction of the model,based on the support vector regression principle,a shortterm traffic flow prediction model based on support vector regression(SVR)has been constructed,and the main impact parameters of the model also have been determined.The principle of whale optimization is introduced.Using whale optimization(WOA),the algorithm optimizes the main influence parameters of the support vector regression(SVR)model,thereby establishing a prediction model based on WOA-SVR.EMD decomposition method is used to decompose the traffic flow data into the sum of different IMF components and residual components.The WOA-SVR model is applied to make predictions and the predicted waveforms of each component is reconstructed.A short-term traffic flow prediction model is established which is based on the EMD decomposition fusion whale optimization(WOA)and support vector regression(SVR)models.In the end,this thesis conducts an example verification and compares the experimental results of different models with actual data.The result shows that the short-term traffic flow prediction model based on EMD decomposition and fusion whale optimization(WOA)and support vector regression(SVR)model has good performance.Compared with the previous SVR model and WOA-SVR model,the accuracy is significantly improved.