| In this uncertain world, the prediction for the future is always the most important, no matter the management of the macro-economic or micro-finance market investment decisions are inseparable from the forecast. But only according to the judgment of the managers, investors and researchers, it is not enough to get an accurate prediction, we must rely on statistical methods or the quantitative models to make quantitative predictions as accurate as possible.From the perspective of econometrics, the development of prediction methods probably go through three stages:prediction model based on ordinary least squares method to estimate the structure of measurement, based on maximum likelihood estimation of time series models as well as non-parametric, non-linear neural network and support vector regression and classification models.The predictive ability of support vector regression and classification mainly from they unique design of the structural risk minimization, that not only minimize the empirical risk (ordinary least squares, maximum likelihood and artificial neural network only done this), fitting ability to maximize the return to the sample, and to maximize the generalization ability of sample forecast final two parts of the balancer. Support vector regression and classification not only as ordinary least squares, maximum likelihood and artificial neural network has the same good fitting ability, and has better predictor of generalization ability.A novel feedback support vector regression algorithm is on the basis of the development of forward support vector regression in the existing literature, the introduction of a feedback loop from the output layer to the input layer of the global feedback loop that can help the model to capture the variables more dynamic features, support vector regression learning have a profound effect on the ability to improve the feedforward, feedback support vector is the greatest contribution of this paper. |
PDF Full Text Request