Exploring the predictive potential of artificial neural networks in conjunction with DEA in railroad performance modeling

This paper is an investigation into the feasibility of using artificial neural networks (ANN) in conjunction with data envelopment analysis (DEA) for performance measurement and prediction modeling of Class I railroads in the United States. For this exploratory study, DEA-ANN are combined into a two-stage modeling approach. While it is frequently used as a benchmarking tool, DEA lacks predictive capabilities. However, ANN has strong nonlinear mapping and adaptive prediction functionality. In this study, the advantages of combining these complementary methods into an integrated performance measurement and prediction model are explored. For this combined approach, a Charnes, Cooper and Rhodes (CCR) DEA model is used to evaluate the efficiency of each decision making unit (DMU) and to capture the efficiency trend of each railroad. Based upon those DEA results, the follow-on backpropagation neural network (BPNN) model predicts an efficiency score and target output for each DMU. This is a new attempt to extend the BPNN model for purposes of best performance prediction. The resulting framework is an effective benchmarking and decision support system which adds adaptive prediction capabilities to current benchmarking practices.