Study On Information Integration And Transportation Demand Prediction Of Regional Traffic

Transportation play an important part in the formation and development of regional economy and proper redistribution of production capability, which connect all sectors of national economy and link the production and consumption together. Transportation is the important insurance for development of production of industry and agriculture and economic activity. It is required for the sustainable development of regional economic environment that transportation governor make right transportation planning and decision-making. In this dissertation we studied several key issues in the regional transportation planning and management, and main contents included as follows.â… . Integration of multi-source regional transport information and construction research of conceptual system. Aiming at the problem of heterogeneity in integration of multi-sources data in area transportation informatization, this paper discussed the integration from two angles that is from data application layer and from semantic layer, and proposed the construction of transportation information platform based on data warehouse according to integration in data layer and application layer to meet the high-level needs of data mining and information value-added. Middleware technology can be applied in regional transportation system's data integration, and XML technology is adopted to realize complex multi-source heterogeneous data integration within the department and different application information systems by making data extraction, transformation and loading, as well as to provide mutual data interface to establish the common data standards and specifications. To solve mapping problem effectively in traffic information systems'semantic integration domain ontology is raised and published and semantic integration mechanism of transportation domain ontology was proposed.â…¡. Study on analysis of regional transport demand association factors. To study the association factors and causality of transport demand, a pre-set range of properties are often required. In order to reduce the subjective effects of human factors and to improve the objectivity of the analysis, the GMDH method based on self-organizing data mining was proposed on the building of regional transport indicator system and using the Granger causality test model for reference to raise a GMDH-based causality model. Combined with association factor analysis model and grey correlation used in the actual project in Yunnan Province that not only for the region's transportation planning and decision-making but also providing an important reference for the follow-up.â…¢. The mid long term forecasts on traffic demand. As transportation demand generated by a variety of factors and in different measure its operation law is quite different, so it is necessary to forecast demand by conducting in different time granularity that are long-term prediction and the volatility analysis. The mid long term regional transport demand forecast study begin with traditional methods of multiple linear regression as a reference as classic GM (1,1) model's flaw in parameter estimation accuracy and improved AGM (1,1) model was proposed, while grey model residuals is a Markov chain, an improved grey-Markov correction model is verified through improved grey Markov model based on historical patterns and internal relations of time series, given the forecast of future volatility of the trend, not only considered from the time series data mining the evolution of the law, but also through the state transition probability matrix of the random response transform to extract the data, the time series data is inherently combine the two properties, a more objective response to fluctuations in demand of credibility.â…£. Volatility study on regional transport demand. Since the transport demand does not increase or decrease monotonically but the volatility and periodicity are always existed. To reduce subjective intervention in the input process of modeling and improve prediction accuracy, polynomial neural network PNN based on GMDH thinking is proposed, the problem of the traditional time series analysis methods not considering the inner influencing factors can be solved, the shortcomings of a variety of factors, according to multi-input of the influencing factors polynomial neural network was generated by self-organized through training and testing and the result with higher accuracy can be get.