Data Driven Prediction Of Urban Traffic Flow Using Fuzzy Methods And The Periodicity Pattern

Since the reform and opening up,the national economic level has increased substantially,which has led to a sharp rise in the number of motor vehicles in China.While the motor vehicle brings convenience to the transportation,it also leads to the deterioration of urban traffic conditions.At present,traffic congestion has become a bottleneck in the development of many cities,so it is urgent to alleviate urban traffic congestion.The short-term traffic flow forecasting estimates the traffic flow in the future by historical traffic flow data,which can effectively improve the travel efficiency of residents and enhance the control of urban traffic by law enforcement personnel.It is an effective means to solve urban traffic congestion problems.However,due to various factors such as weather and air pollution,there is a high degree of uncertainty and randomness in historical traffic flow data,which makes accurate and reasonable prediction very difficult.For the accurate prediction of urban short-term traffic flow,the data driven fuzzy prediction method considering the periodicity pattern is studied and applied for short-term traffic flow prediction in this thesis.The main work of this thesis is as follows:First at all,an overview of the current mainstream traffic flow prediction methods is presented.The characteristics of urban traffic flow prediction are introduced firstly.Next,five mainstream traffic flow prediction approaches are introduced.And then,the above five traffic flow prediction methods are analyzed and compared.Finally,a summary is made.Subsequently,in order to reduce the influence of uncertainty in historical traffic flow data and to improve the accuracy of urban short-term traffic flow prediction,one Adaptive-network-based fuzzy inference system model(ANFIS)based hybrid model which take into account the periodicity pattern is proposed for urban traffic flow prediction.The model combines the periodicity pattern of the traffic flow and the data-driven traffic flow prediction model to generate the final prediction results.First at all,the periodicity pattern of traffic flow data is extracted,and the periodicity pattern is removed from the original traffic flow data to obtain the residual data which is utilized to train the traffic flow prediction model.Finally,the output of model and the periodicity pattern are combined to obtain the final traffic flow prediction result.The ANFIS is used as the data-driven traffic flow prediction model of the proposed approach.To verify the superiority of the proposed method for urban traffic flow prediction,two different traffic flow prediction experiments are carried out.The experimental results show that the proposed hybrid model has obtained the best prediction results in different short-term prediction experiments,which proves the effectiveness of the proposed method.Finally,in order to solve the traffic flow prediction input variable determination problem and the rule explosion problem when the fuzzy method has many input variables,a functionally weighted single-input-rule-modules connected fuzzy inference system(FWSIRM-FIS)based hybrid model for urban traffic flow prediction is presented.The model combines the periodicity pattern extraction method with the FWSIRM-FIS model to generate the final result for the urban traffic flow prediction.First of all,the traffic flow periodicity is extracted from the historical traffic flow data,and the corresponding residual data is obtained by removing the extracted pattern from the historical data.Then the FWSIRM-FIS model is trained after determining the optimal input variables which are chosen by the partial autocorrelation analysis(PACF)technique.Finally,the output of the FWSIRM-FIS model is combined with the periodicity pattern to generate the final prediction result.In order to verify the effectiveness of the hybrid model for urban traffic flow prediction,it is compared with three traffic flow prediction methods: ANFIS model,BPNN model and pruned fuzzy inference system(PFLFIS).Experimental results show that the model can obtain the best prediction performance in different urban short-term traffic flow prediction experiments,which proves the effectiveness of the hybrid traffic flow prediction model again.