Research On Prediction Of Traffic Volume At Non-detector Intersections Based On Genetic Algorithm And Stepwise Regression Analysis

As an important research area of the Intelligent Transportation System (ITS), Traffic Guidance System is currently recognized the best way to improve transport efficiency and mobility. Traffic flow guidance function which in order to implement is to obtain accurate real-time traffic foundation information as the basis. This paper is based on the national 863 project " key technology research of a new generation of intelligent traffic control system " neutron subject " large-scale travel demand forecasting technology", mainly to resolve the forecast accuracy of traffic volume of intersections without detector ,which is in order to provide accurate and real-time traffic information for the large-scale traffic control and traffic guidance.Whether traffic flow control or traffic flow guidance , we need the accurate and real-time traffic flow data as a basis.With the basic data of traffic volume, we need to continue looking for the higher accuracy method to predict the volume. Though the methods to predict the volume of intersections without detectors is varied and many methods have been widely applied to the practical, the predicted result still can not satisfy traffic flow control and traffic flow guidance demand.in particular we need to improve the accuracy of forecast.It is necessary for us to have further study with forecasting methods of volume of intersections without detector.The springboard of article is the traffic flow forecasting solution in the practical problems of the dynamic traffic flow guidance, committed to the real-time and dynamic forecasting volume of intersections without detector for the entire road network .This paper is on providing high-quality forecast for the volume of traffic data preprocessing and detector intersections without traffic prediction method for the start.The core of article is the model and algorithm of the forecasting method.It is a focus of raising the prediction accuracy in the paper.The full text is divided into five chapters: Chapter 1: IntroductionFirst introduce the status and research significance of traffic flow forecast on, a comprehensive summary and review of the main traffic flow forecasting models and algorithms. Then put forward traffic forecasts problem. At last, the purpose and meaning of this research, the main research content and the composing of the whole paper are presented.Chapter 2: Research on the pre-processing technology of the traffic volume First introduced the technology and collection form of date acquisition, anb introduce detector types and traffic information collection methods. And analyse the characteristics of traffic volume , pointed out that the traffic volume has relevance and similarity of weeks. Then summarize and presente data preprocessing technology, including the wrong data, missing data, inaccurate data identification and restoration techniques and data smoothing methods. Finally,we process the date of Changchun detectors and processed data show that the method can improve the quality of data and ensure the model input data accurate.Chapter 3: introduce the conventional research on the volume forecast of intersections without detectors.Introduce the principal component analysis and cluster analysis which are most used in statistical data processing, details of the principal component analysis and cluster analysis how to use in foreasting the traffic flow of intersections without detectors.then most analyse the advantages and disadvantages of the principal component analysis and cluster analysis in foreasting the traffic flow of intersections without detectors, and finally proposes to construct a new method of thinking of forecasting the traffic flow of intersections without detectors .Chapter 4: Research on the volume forecast of intersections without detectors based on genetic algorithm and stepwise regression analysis.This chapter first analyses the purpose of the study of volume forecast of intersections without detectors. Then introduce the stepwise regression analysis of regression analysise in detail, pointed out the shortcomings of this method: stepwise regression analysis can only identify a subset of regression, not considering optimal criteria, only available from several controversial subset the IV opportunity to choose and not to eliminate variables between the multicollinearity, in essence, that it is not the best criteria for the optimal variables Subset.Introduce in the advantage of genetic algorithm optimization and propose the algorithm of the volume forecast of intersections without detectors based on genetic algorithm, final validate the algorithm with the data of Changchun intersection , contrast to the predicted results of principal component analysis and cluster analysis certificating that the feasibility of the method and higher precision.Chapter 5: Looking summaryThis paper presented the results achieved and the shortcomings ,pointe out the direction and goals of the future work.