Data Envelopment Analysis (DEA) has been applied in many studies in the banking sector. Conventional DEA models consider the system as a single-process black box. There are however a number of so-called network DEA approach that consider the system as composed by distinct processes or stages, each one with its own inputs and outputs and with intermediate flows among the stages. In this paper some of these approaches that have been applied to banking sector are reviewed and in particular applications of two-stage DEA to Iranian banking branches are presented and discussed. The conclusions of this research are that two-stage DEA models have a greater discrimination power than conventional, single-process DEA and since they allow a more fine-grained analysis their results are generally more valid and useful. The main drawback is the need for more detailed data (i.e. at the process level) and the greater complexity of the resulting models, especially if there are inputs or outputs that are shared among the processes.

Carbon-capture-and-storage (CCS) is an important technology to reduce CO$_2$ emissions. A commercial method to establish the CCS on a large scale is to sequestrate CO$_2$ in depleted oil reservoirs and to combine it with enhanced oil recovery (EOR) operations. In this way, the CO$_2$ emission is reduced, as well as the oil production increases. The joint CCS-EOR planning problem determines the optimum allocation of existing CO$_2$  to depleted reservoirs and the scheduling of the EOR operations. This paper presents a deterministic MIP model which is a modification of an existing model in the literature. Then, this model is extended to a two-stage stochastic model in which the parameters expressing the initial oil yields and the periodic depletion factor of oil yields associated with reservoirs are uncertain, and the uncertainty is realized as soon as the operation of the reservoir is started. Our stochastic model is computationally more efficient than the existing model in the literature, due to the reduction of binary variables, as well as the absence of “non-anticipativity constraints”. Instead, our stochastic model is less realistic. The proposed models are examined over two case studies taken from the literature. The obtained results confirm the higher effectiveness of our stochastic model.

This study considers instant decision-making needs of the automobile manufactures for resequencing vehicles before finalassembly (FA). We propose a rule-based two-stage stochastic model to determine the number of spare vehicles that shouldbe kept in the pre-assembly buffer to restore the altered sequence due to paint defects and upstream department constraints. First stage of the model decides the spare vehicle quantities, where the second stage model recovers the scrambledsequence respect to pre-defined rules. The problem is solved by sample average approximation (SAA) algorithm. Weconduct a numerical study to compare the solutions of heuristic model with optimal ones and provide following insights: (i) as the mismatch between paint entrance and scheduled sequence decreases, the rule-based heuristic model recovers thescrambled sequence as good as the optimal resequencing model, (ii) the rule-based model is more sensitive to the mismatchbetween the paint entrance and scheduled sequences for recovering the scrambled sequence, (iii) as the defect rateincreases, the difference in recovery effectiveness between rule-based heuristic and optimal solutions increases, (iv) asbuffer capacity increases, the recovery effectiveness of the optimization model outperforms heuristic model, (v) asexpected the rule-based model holds more inventory than the optimization model.