Performance measurement for network DEA with undesirable factors

Traditional Data Envelopment Analysis (DEA) models take Decision-Making Unit (DMU) as a 'black box' without considering the inputs/outputs of its intermediate production processes. To provide efficiency enhancing information regarding the sources of DMUs' inefficiencies, researchers investigate network DEA models. This paper proposes a network DEA model in the presence of undesirable factors. In the network DEA model, a DMU is composed of a set of interdependent sub-DMUs, that is, input of a sub-DMU may be an undesirable output of another sub-DMU. We develop a method of estimating efficiency of such DMU, and analyses efficiency relationship between a DMU and its sub-DMUs. Our model provides a way of improving performance of a DMU through identifying its inefficient sub-DMUs. Numerical examples are used to show our results.