Although the error analysis and model validation tests presented
in Section ‘Error analysis and model validation’ is meaningful
to compare the performance of the various logistic modelling
approaches used in this study. However, from a statistical
perspective, test proposed by Diebold–Mariano (DM) [35] is used
to analyze the comparative performance of the logistic and
logistic-population approaches used in this study. In DM test, null
hypothesis assumes that forecasting accuracy of two models is
same and vice versa is assumed as the alternative hypothesis.
Mean square error (MSE) is used as a loss function. Thus, DM statistics
can be described as Eq. (8):