Variability and relationships: Apart from the changes in the
company's structure or its atmosphere, the consideration of
assumed stable relationships among a model's variable is very
important. With the supposition of complete data availability,
the MLPNN models are more suitable for explaining the
complex non-linear relationships between variables. The
SARIMA-MLR model assume linear relationship among variables. But, there is also a possibility that the SARIMA-MLR
model would reject an independent variable at mean which
may be significant at other quantiles. Unlike the SARIMA-MLR
and MLPNN models, the SARIMA-QR model provides more
information about relationships between variables at different
quantiles. In case of poor knowledge of relationships, the
seasonal naïve forecasting model can be chosen.