Traffic Flow Prediction of Chaos Time Series by Using Subtractive Clustering for Fuzzy Neural Network Modeling
Pang Ming-bao, Xinping Zhao · 2008
The method was studied about traffic flow prediction by using subtractive clustering for fuzzy neural network model of phase-space reconstruction. The prediction model of traffic flow must be established to satisfy the intelligent need of high precision through analyzing problems of the existing predicting methods in chaos traffic flow time series and the demand of uncertain traffic system. Based on the powerful nonlinear mapping ability of neural network and the characteristics of fuzzy logic, which can combine the prior knowledge with fuzzy rules, the knowledge base of the traffic flow predicting system was established by using fuzzy neural network model based on subtractive clustering. Subtractive clustering generates the number of fuzzy rules and the clustering centers are regarded as the initial training parameters of the predicting modeling. The predicting model of fuzzy neural network can be quickly trained online. Genetic algorithm was used in determining the clustering radius. The simulation result shows its correctness and feasibility.