Data availability: Traditional time series (ARIMA and SARIMA),
hybrid seasonal time series (SARIMA-MLR and SARIMA-QR)
and MLPNN models proportionately perform better based on
the length of historical data (both response and predictor
variables). More and detailed the data, superior the accuracy.
Comparatively, the MLPNN models require higher amount of
data than the time series models for better approximation, i.e.,
the predictions are less reliable and accurate when the training data are insufficient. On the other hand, the seasonal naïve
forecasting model is more suitable when the data of response
and predictors variables are qualitative