Introducing Prediction Concept into Data Envelopment Analysis Using Classifier in Economic Forecast

Using data envelopment analysis (DEA) to evaluate the relative efficiency of decision-making units (DMUs) needs pre-determining all the input and output variables. However, this may not be the case in the practical applications. This paper proposes a DEA-Classifier model to predict the efficiency level of DMUs with partial input and output variables. This model first divides the historical DMUs into two classes (high efficiency and low efficiency) based on their efficiency scores determined by DEA and then uses the machine learning method to establish a mapping relationship between the class labels and the partial input and output variables. For a new DMU, the trained classifier can predict whether it belongs to a high or low efficiency class using its partial input and output variables. The proposed DEA-classifier model is validated by theoretical analysis, simulation data, and real-life datasets. The experiment results indicate that the proposed method has good adaptability and it can be applied in many areas including economic forecast and prediction.