Uncertainty: The first problem of SARIMA-MLR and MLPNN
models is that they only provide point forecasts. These models
produce only mean forecast, i.e., the forecast will provide only
50% of service level. In addition to this, the SARIMA-MLR model
assumes Gaussian distribution for high volatile and skewed
time series. As the true distribution of demand or sales is not
normal, the extrapolation of higher quantiles from the mean
forecast is not going to reflect the reality (Khosravi et al., 2011).