Uranium Resource Price Forecasting Based On Chaos Theory And Support Vector Regression

As the material basis for the development of nuclear power, the changing rule of theuranium resources is influenced by factors such as economic, political. Study the tendency ofthe international uranium resource prices is important to making plans on development andutilization of nuclear energy. At present, the price prediction research is mainly based on timeseries, the researching methods mainly divided into two categories. One is traditional method,such as ARMA model, GARCH model and so on; another is artificial intelligence method,such as neural network model, support vector model, fuzzy regression model and so on.Although the neural network has the character of non-linear, in the application it affected bythe initial value, slowing training speed and easy to fall into local minima, and sometimes itoccurs over-learning or generalization ability phenomenon, it will affects the stability of themodel and the accuracy of forecasting.Followed the structural risk minimization prince, the SVM can solve some problemsbetter, like the small sample, nonlinear, high-dimension, local minimum and so on. It solvedthe dimension disaster, and it is a new researching hot-point. The paper mainly includes thefollowing aspects:(1) Introduced the Chaos theory and the Markov decision thought, calculating theLyapunov exponent to explain the chaotic of the time series, using the Cao method to decidethe smallest embedding dimension, and reconstruct the time series of uranium resources(2) This article describes the algorithm, structure and model of support vector machine,introduce the type of the kernel function, which the selection of parameters affectperformance and forecast results of support vector machine model. In this paper, using thetheory of particle swarm algorithm to optimize the parameters of support vector machinemodel. (3) On the basis of the reconstruction of the uranium resources price time series, builtthe uranium resources price prediction model based on RBF neural network, SVR andPSO-SVR, and then comparing the simulation experiment.The results show that the prediction model based on PSO-SVR has the higher accuracyof the forecast results, verifying the applicability and effectiveness of the internationaluranium resource prices based on PSO-SVR model.