Multivariate statistic combination forecasting model of wine quality
According to [31], the individual multivariate statistical methods include principal component regression method (PCR),
partial least squares regression method (PLSR), and modified partial least squares regression method (MPLSR) [23]. In order
to reflect the prediction accuracy, confidence interval (CI) with 99% confidence is presented. The prediction results of some
popular multivariate statistic methods are shown in Table 3.
Table 4 shows the evaluation indices (SSE and MSE) of multivariate statistical methods. It is evident that MPLSR has a
relatively high accuracy in prediction.
The optimal combination forecasting models based on MESS, MSAE and MMAE are built, and the prediction equations of
combination forecasting model (2), (4), (5) are: