A Forecasting Model with Consistent Adjustments for Anticipated Future Variations

Due to the limitation of most statistical forecasting models ignoring contextual information, judgmental adjustment is a widespread practice in business. However, judgmental adjustment still suffers with many kinds of biases and inconsistency inherent in subjective judgment. Our approach uses an adjustment mechanism concerning only with critical cue factors evaluated with genetic algorithm to alleviate problems caused by collinearity and insignificant sporadic variables usually arising in least square type estimators, and to derive more realistic parameter estimation. In case there are anticipated variations in the forecasting horizon and can’t be handled by the model alone, this adjusting mechanism, formulated in a set of equations, can be used to assess mixed effect of cue factors consistently and effectively without subjective judgment involved. Empirical results reveal that this adjustment mechanism could significantly reduce MAPE of forecasts across seasons with improvement mainly coming from large size adjustments.