Developing a hybrid comparative optimization model for short-term forecasting: an ‘idle time interval’ roadmap for operational units’ strategic planning

Data drain and data uncertainties for rival units affect the reliability and effectiveness of strategic plans for individual operational units. This study introduces a stochastic, multi-stage, optimization technique for short-term forecasting that intends to assist policy makers in developing ‘flawless’ plans for their organizations during the idle time interval in which official data and balance-sheet reports of the competitors are unavailable. The developed technique, called SDEANN, draws on the ‘deterministic’ data envelopment analysis (DEA) method, ‘regression-type’ artificial neural networks (ANNs), and the contamination of the outputs of the DEA analysis with statistical noise. Statistical noise represents the bias of a ‘deterministic’ sample optimum production frontier when generalization or the uncertainty of the data used becomes the issue. The SDEANN model respects the monotonicity assumption that prevails in microeconomic theory, uses the DEA definition of efficiency, and addresses the dimensionality issues of ANNs with minimum sample size requirements.