Abstract: A GM(grey model)(1,1)-RBF (radial basis function) model based on RAGA(real coded accelerating genetic algorithm) has been established. There exist conspicuously systematical deviations when we are fitting the data using the traditional GM(1,1) model. But the shortcoming has been overcome by the new model. The model has the following advantages: Firstly, it can hold the certainty of the data; what ' s more, the advantages in the uncertainty domain in neural network are interfused. The predicted results indicated that it is more precise than the traditional methods. The scientific rationality of portfolio forecast model used for medium-and long-term forecast respectively is verified . The result will provide a reference in making policy.

Key words: RAGA , GM(1,1) , radial basis function neural network,  portfolio prediction , Shanxi Province