Stable Computation of Least Squares Problems of the OGM(1,N) Model and Short-Term Traffic Flow Prediction

Abstract

The optimized grey multi-variable model, used to overcome the defects of the grey multi-variable model, is studied. Although this model represents a substantial improvement of the grey multi-variable one, unstable computation of the grey coefficients arising in ill-posed problems, may essentially diminish the model accuracy. Therefore, in the case of ill-posedness we employ regularization methods and use the generalized cross validation method to determine the regularization parameters. The methods developed are applied to the urban road short-term traffic flow prediction problem. Numerical simulations show that the methods proposed are highly accurate and outperform the grey multi-variate, the autoregressive integrated moving average, and the back propagation neural network models.