Study Of Combination Forecast And Application In Traffic Flow Forecast

With the development of national economy, the people's demand of transportation becomes increasingly expanding. The contradiction between the growth of traffic volume and the road condition is becoming prominent, which has greatly restricted the development of the social economy. Accurate real-time prediction of short-term traffic flow is the key technology in traffic control and guidance. In theory and practical applications indicate that combination forecasting method has higher accuracy of the forecast than single prediction method, enhances stability, and has a higher ability to adapt to the future and changing predict climate. The dissertation researched on the combination forecasting methods and application of the model. The main results achieved in this dissertation can be summed up as follows:a) Existing combination forecasting methods are classified, combination forecasting methods and single prediction methods are compared in accuracy, then, the reasons that combination forecasting methods bring the error are analyzed.b) The optimal combination forecasting model are build on the square sum of the error as the objective function .Then the model is to be solved, and we get the right formula for calculating the weight coefficient and the square sum of the error expression; We also provethat when error information matrix E_m∈Z^m,m ,the weight coefficients of optimalcombination forecasting methods are positive; We estimate the range of the square sum of the error in combination forecasting more accurately, and give the conditions that simple average method become the optimal combination forecasting method.c) On the basis of variance reciprocal weighted method, error reciprocal variable weight combined forecasting method is advanced. In addition, we use polynomial function that the time t is as the variable to approach variable weight coefficient, solve the coefficient matrix of the polynomial in iterative, and prove that the convergence of iterative algorithm, in the end, we get a new solution method of the weight coefficient in variable weight combined forecast.d) Combining with the measured data, we predict traffic flow by four individual prediction models and the optimal weighted combination forecasting model, and also analyses compare the individual prediction model and optimal weighted combination of forecast model. As a result, it's proved that combination forecasting method would improve prediction accuracy.