S. Benítez Peña, P. Bogetoft, D. Romero Morales
Here we present an integrative approach to Feature Selection (FS) in Data Envelopment Analysis (DEA) for Benchmarking. DEA aims at benchmarking the performance of decision marking units (DMUs), which use the same inputs and produce the same outputs, against each other. In DEA, the efficiency of DMUs is measured as the weighted summation of the outputs divided by the weighted summation of the inputs, and the weights are found solving a Linear Programming problem for each DMU. DMUs with a score equal to one are in the so-called efficient frontier. FS for both outputs and inputs has a significant impact on the shape of the efficient frontier, as well as the insights given to the inefficient DMUs. Here, the DEA model is enriched with zero-one decision variables modeling the selection of features, yielding a Mixed Integer Linear Programming formulation. Numerical results will be presented, highlighting the advantages that our single-model approach provide to the user.
Keywords: Mixed-Integer Programming, Feature Selection, Data Envelopment Analysis
Scheduled
GT4-2 Multivariate Analysis and Classification
September 3, 2019 4:50 PM
I3L10. Georgina Blanes building