Forecasting the Confidence Interval of Efficiency in Fuzzy DEA

Abstract

Data envelopment analysis (DEA) is a well-known method that based on inputs and outputs calculates the efficiency of decision-making units (DMUs). Comparing the efficiency and ranking of DMUs in different periods lets the decision-makers prevent any loss in the productivity of units and improve the production planning. Despite the merits of DEA models, they are not able to forecast the efficiency of future periods with known input/output records of the DMUs. With this end in view, this study aims at proposing a forecasting algorithm with a 95% confidence interval to generate fuzzy data sets for future periods. Moreover, managers' opinions are inserted in the proposed forecasting model. Equipped with the forecasted data sets and concerning the data sets from earlier periods, this model can rightly forecast the efficiency of the future periods. The proposed procedure also employs the simple geometric mean to discriminate between efficient units. Examples from a real case including 20 auto-mobile firms show the applicability of the proposed algorithm. (original abstract)