Short-term Traffic Flow Forecasting Research And Application Based On Clustered-WNN-ARIMA

Recently, along with the rapid increase of national GDP citizens'income and the development of progression of urbanization, the number of vehicles increases daily, as the basic infrastructure which supports the social economic activities, boosted up peoples'life level, but the problems that the urban transportation system shows such as energy consumption, urban traffic congestion and traffic safety problems becomes much more serious than ever, and it has constrained the further development of our transportation industry. Due to the features of uncertainty, nonlinear and time variation of transportation system, traditional signal control system can not effectively solve the previous problems and is not fit for the large city such as Beijing, Shanghai and so on. The ITS (Intelligent Transportation System) has been proposed and can theoretically solve previous problems. The ITS build a novel transportation control system using new developed information fusion technology to reduce traffic jam, energy and pollution. The traffic flow prediction is a very important component within the components of ITS and is one of the key points of ITS.However, urban road transportation system is a time varied, extremely complex system which is vary hardly predicable, and short-term traffic flow prediction model or method has direct relation with the effect of ITS. Short-term traffic flow prediction is the main research area of traffic state prediction, short-term traffic predictions are necessary inputs for advanced traffic management systems (ATMS) and advanced traveler information systems (ATIS) and will powerfully make sure that effective running of transportation system. Short-term traffic flow information shows the features of self similarity which makes it predictable, and high dimension, nonlinear and time variation which make it harder to accurate and timely forecasting. Finally, based on the deeply study on the short-term traffic flow forecasting problem and the researches of exist short-term traffic flow forecasting models and methods, we proposed a improved short-term traffic flow model, and improved the effect of the traffic control strategies that developed by our team. The main contributions of this paper as follows:(1) Based on the researches of recently short-term traffic flow forecasting models and analysis the basic features'variables and characteristics of traffic flow in different traffic environment, using the preprocess methods process the error data and missing data of collected traffic flow to reduce the negative effect on the accuracy of traffic flow forecasting model.(2) Based on the basic WNN-ARIMA model and combined with clustering analysis technology, we proposed a new novel short-term traffic flow forecasting model called clustered based WNN-ARIMA short-term traffic flow forecasting model, and tested the model using the data downloaded from Open Traffic Database-PeMS and the traffic simulation platform software VISSIM, the results of experiments shows that a higher and stable accuracy has been achived.(3), Based on the short-term traffic flow forecasting model, we improved the effect of existing traffic signal control system we developed, and tested it with the road network including the 19 main intersections of DaLian which built with VISSIM, and achieved expected results.