Evaluating the efficiency using data envelopment analysis(DEA) and machine learning

Abstract

Data Envelopment Analysis (DEA), a non-parametric productivity analysis, has become an accepted approach for assessing efficiency in a wide range of fields. Despite of its extensive applications, some features of DEA remain unexploited. DEA is good at estimating "relative" efficiency of a DMU, it can tell us how well we are doing compared to our peers but not compared to a "theoretical maximum." Thus, to measure of efficiency of new DMU we have to develop DEA previously at same method with the data of used DMU. Also we can not predict the efficient level of new DMU. Second, because DMUs are directly compared against a peer or combination of peers. DEA offers no guide where relatively inefficient DMU improve. Third, DEA identifies peer DMUs and targets for inefficient DMUs, but it doesn``t provide the stepwise path for improving the efficiency of each inefficient DMU considering the difference of efficiency. In order to overcome this limitation of DEA, I suggest to a new methodology. The methodology I proposed is a hybrid analysis utilizing the machine learning and DEA, and it can be summarized as follows. First, I apply a DEA to evaluate the efficiency of the DMUs with their multidimensional inputs and outputs. After that, I clustered the DMUs together through the tier analysis, which recursively apply the DEA to the remaining inefficient DMUs, and then generated the DMU classification rules using the C4.5, the decision tree classifier, with the DMU tiers that had identified by the tier analysis. Second, DEA offers no guidelines where relatively inefficient DMUs improve since a reference set of an inefficient DMU consists of several efficient DMUs. Hence, I utilized a self-organizing map (SOM), which is one of clustering tools for grouping similar DMUs according to inputs, for the inefficient DMU to select one efficient DMU in a reference set as a benchmarking target. With the tiers identified by the tier analysis, it could provide the guidelines for stepwi...