An objective-free outlying point detection model in data envelopment analysis

Data envelopment analysis (DEA) is a mathematical programming approach for benchmarking. Using extreme observations to identify superior performance makes DEA vulnerable to outliers. While there are many studies on detecting outliers for DEA, most focus on specific applications and objectives, e.g. orientation, and thus have limitations. We address the limits of conventional objective-dependent approaches and the need for an objective-free outlier detection mechanism, and propose an outlier detecting method that requires no pre-specified objectives. In addition to allowing for the identification of outliers, the method described is consistent with a relaxed set of DEA axioms.