where xt is the measured value for the tth samples, xˆt is the model predicted response values for the tth sample, and N is
the number of verification samples, respectively.
According to Eq. (7), we obtain the SSE’s and MSE’s of 8 combination forecasting models and 2 individual forecasting
models in Table 2, respectively. It is evident to see that the weighted coefficient vector ω = [0.3406, 0.6594] is the optimal
one. Therefore, in this example, the combined forecasting model based on the criterion of minimizing the sum of absolute
error (MSAE) performs the best