Forecast approach using neural network adaptation to support vector regression grey model and generalized auto-regressive conditional heteroscedasticity

Triangular fuzzy series forecasting based on grey model and neural network

“…After using the BP network to predict the residual series, we can get the modified values of Sˆ, F and M x from (17). According to (18), the predicted triangular fuzzy series of NNTFGM (1, 1) is obtained. The results are shown in Table 3.…”

“…This characteristic promotes the extensive application of the BP neural network. The applications of the BP neural network have shown that it is suitable for the prediction of the fluctuating series and has good effectiveness [17][18][19][20][21][22].…”

Dynamic grey platform for efficient forecasting management

“…Many scholars have adopted the grey theory for analysis and decision making in situations where enough data is not available to be effectively used by traditional prediction approaches. The GM(1, 1) model of grey theory is commonly used for forecasting in several application fields such as pollution and energy [6], social management [7], industrial engineering [8], financial management [9,10] and logistic transportation [11].…”

Time Series Forecasting Model Based on Discrete Grey LS-SVM

“…On the other hand, the robustness of the least square method is poor, which affects the stability of the model. LS-SVM is a theory based on SRM, which has very good generalization ability [19]. LS-SVM method does not need to require the prediction variables obey normal distribution.…”

“…The moving average autoregressive exogenous prediction model is combined with grey predictors for time series prediction [18], and it is proved that the hybrid method has a greater forecasting accuracy than the GM(1,1) method. Another study introduces a support vector regression grey model (SVRGM) which combines support vector regression (SVR) learning algorithm and grey system theory to obtain a better approach to time series prediction [19]. Both the grey model and the SVM model do not need to know whether the prediction variables obey normal distribution, do not to require too much statistic sample.…”