Developing a new integrated artificial immune system and fuzzy non-discretionary DEA approach

Abstract

As a useful performance evaluation and decision-making tool, data envelopment analysis (DEA) has been proven to be an excellent data-oriented efficiency analysis method when there are multiple inputs and outputs. However, when working with large datasets, DEA requires more time to solve and calculate the optimal values for each decision-making unit (DMU). To close this methodological gap, this study proposes a new integrated fuzzy nondiscretionary DEA (FNDEA) model and artificial immune system (AIS) to predict and find the optimal values of DMUs. In so doing, we first modify an FNDEA model to classify the set of all DMUs into efficient and inefficient, and immune system (AIS) is used to predict and find the optimal values of DMUs. Then, a modified fuzzy nondiscretionary additive DEA model, which is designed as a middle chain, is used for sensitivity analysis of inefficient DMUs to determine their target outputs, which are called antigens. Finally, we try to reduce the distance between these two steps in order to predict the optimal values (with nearest distance to antigens) of inefficient DMUs and improve their efficiency by using a combined AIS and FNDEA model called the FNDEA–AIS approach. To illustrate the advantages of the proposed FNDEA–AIS approach, a dataset from 24 Iranian forest management units is collected; the results indicated that our new FNDEA–AIS approach (in comparison with other well-established performance prediction techniques) exhibits better convergent validity and high correlation with low error rate to predict optimal values of inefficient DMUs. The main conclusion is that our new applied method, developed for the first time herein, provides a suitable improvement over previously developed methods, such as artificial neural networks (ANNs), to detect inefficiency and to improve the overall performance when analyzing different types of real-life problems.