Time Series Forecasting Using a Hybrid Principal Component Analysis and Global Ridge-Regression Neural Network Model

This research of forecasting the most developed is the time series, which uses a quantitative approach to the data of the past that made reference to forecasting the future. This experiment propose methods to complete time series by reducing the PCA features and improvements Radial Basis Function Neural Network (RBFNN). Where RBFNN is done by combining the functions of RBF and weighting function to produce optimal outcomes. The weights are gotten from regression. Global-ridge regression adds a regulation to give the optimal parameters that produce an optimal weights output. While the feature improvements by using PCA. In the application of research on forecasting the data of SMEs that have a variable with a value of nonlinear show that the method of incorporation RBFNN and global-ridge regression is called (GRNN) have more accurate results compared to some of the methods shown in the value of MAPE, and MSE of all variables data.