Analysis And Application Of Dynamic Traffic Data Based On Granular Computing

All of the research on the traffic problems is based on the traffic data. Especially, to the intelligent transportation system, the analysis and processing to dynamic traffic data is very important. With the development of the different kind data acquisition equipment and com-munication technology, the quality and quantity of the traffic data is improved for obtaining the traffic condition and solving the traffic problems. The modern traffic data analysis tech-nology should have three features for processing the massive, multi-source and heterogene-ous traffic data. One of the features is to fusion the traffic data for avoiding the data wasting. Another one is to mine the traffic data to find the internal relations of the different data for acquiring the transportation conditions. The last one is to analysis the traffic data quickly to satisfy the requirement of traffic control.The Granular Computing, GC, is a new issue of the intelligent data analysis. The theory uses granular to define, measure and reasoning the research objects. It can treat the problems from different aspects and levels. Because the GC can't fall into local minimum, it is good at processing the massive, fuzzy, uncertain and incomplete data. Now, the GC theory is use for pattern recognition, image process, intelligent searching and so on. But until now, it is still not used to process the traffic data.With the analysis to the feature of the dynamic traffic data and the GC, and the demand of the modern intelligent transportation management system, the GC is used to process the dynamic traffic data(traffic flow, traffic average speed). The thesis constructed traffic granu-lar computing theory system and introduced the model building. Taking the short-term traffic flow forecasting, traffic congestion detection and recognizing the discrepant traffic data for examples, the thesis introduced how to use the traffic granular computing.Main contents contain:(1) Research on the GC Theory. This part introduces how the theory comes and devel-ops and express that the development of the theory is natural trend as the methodology and the practical application angle. The theoretical frame and application area of the theory that contain the basic concepts, basic problems and relevant mathematic tools are introduced. At the end of the part, basic model and algorithm of the theory are introduced. (2) Traffic granular computing theoretical frame and model building. In this part, the GC is introduced into the dynamic traffic data processing area. The whole traffic granular computing theoretical frame is constructed by inheriting and developing the GC. On the one side, the traffic granular computing inherits the basic idea and substantive characteristics of the GC. Another side, the traffic granular computing fusions the existing good method for the traffic data processing area and the GC. The part also proposes the thought and principle of model building based on the traffic granular computing.(3) Short-term traffic flow forecasting based on the traffic granular computing. A short-term traffic flow forecasting model is built based on the traffic granular computing. In this model, the rough set and neural network are the tools to conduct and compute the granular. Firstly, traffic data pre-process granular are built by the rough set, so the attribute and quan-tity of the traffic data are reduced and the relationships between the data are extracted. Sec-ondly, the relationships can be used to construct the short-term traffic flow forecasting gran-ular based on neural network.(4)Traffic congestion detection model based on the traffic granular computing. In this part, the traffic granular computing is used to detect the traffic congestion on the city road network. The granular are constructed by fuzzy quotient space theory. Firstly, the impact factor evaluate system of traffic congestion on the road network is constructed. Secondly, the input data set of the model are normalized at different levels and the roads on the city net-work are cluster analysed by choosing the appropriate granular. At the end of the part, the congestion road are determined by the cluster analysis results that are processed at different levels.(5)Recognizing discrepant traffic data based on granular computing. In this part, Dis-crepant Traffic Data are recognized by the model based on the coupling of rough set and support vector machine. Firstly, the attribute and the quantity of the traffic data are facilitated and the input set of forecasted model are determined. Secondly, the Least Square is con-ducted to the SVM that is use to recognize the discrepant traffic data.