This study developed the ANN model using ‘IBM SPSS Statistics
21.0’ to predict the predicted error that has not been sufficiently
explained by the SARIMA model. To improve the prediction accuracy
of the ANN model, it is important to determine independent
variables. This study used the following independent variables: the
observed values at time t 12 (Zt-12), the predicted value by using
the SARIMA model at time t ðbL
tÞ, the predicted value by using
SARIMA model at time t-12 ðbL
t12Þ, the predicted error by using
SARIMA model at time t ðest
Þ, and the predicted error by using
SARIMA model at time t-12 ðest
12
Þ. These independent variables
were established by considering that the time series data of electricity
consumption by educational facility would have seasonality
of one cycle per year