data envelopment analysis-based machines (deam)

Data envelopment analysis (dea) is a nonparametric method that aims to use scientific methods in order to investigate the performance of decision-making unit (dmu). one of the interesting subjects in dea is the minimization of the empirical error, satisfying, at the same time, some shape constraints (convexity and free disposability). unfortunately, by construction, dea is a descriptive approach that is not concerned about preventing overfitting. in this research, we introduce a new approach that allows for assessment polyhedral technologies following the structural risk minimization (srm) form. this method is called data envelopment analysis-based machines (deam). it must be admitted that the new approach controls the generalization error of the method, the corresponding estimate of the method does not suffer from overfitting. also, the concept of ε-insensitivity is also advanced, generating a modern and more robust explanation of technical efficiency. accordingly, we present that deam can be seen as a machine learning-type development of dea, favorable the similar microeconomic postulates except for minimal extrapolation. eventually, the performance of deam is estimated through simulations. it goes without saying that the frontier estimator derived from deam is greater than that associated with dea. the error of obtained from deam are smaller in all the schemes analyzed, careless of the number of variables and dmus.