Variable selection in data envelopment analysis: A random forest approach with data augmentation

This study aimed to measure the efficiency scores of 75 dairy buffalo farms in the province of Nueva Ecija, Central Luzon, Philippines, using an input-oriented, variable-return-to-scale Data Envelopment Analysis (DEA) model. The farmer-informants or decision-making units (DMUs) were categorized as smallholders, family modules, and semi-commercial in operations. Personal interviews using structured questionnaires were done to gather various information on the socio-economic and management practices of the DMUs. Output in the form of volume and value of milk produced and inputs such as quantities and costs of biologics, feeds, forage, and labor were also collected and evaluated among individual DMUs. The efficiency scores were computed using PIM-DEA software, which identified fully efficient DMUs lying on the frontier line (scores of 1.0) and those enveloped by it (inefficient DMUs with scores of less than 1.0). The overall mean Technical Efficiency (TE), Allocative Efficiency (AE), and Economic Efficiency (EE) scores among the DMUs were 0.80, 0.81, and 0.65, respectively. Most of the inefficient DMUs were in the smallholder category. In sum, smallholder DMUs classified under low and moderate TE clusters should reduce their inputs by 53.31% and 40.01%, respectively, to become fully efficient. Likewise, higher lambda values among efficient peer DMUs indicate the best practice frontiers that the inefficient peer DMUs can benchmark with. Extension and advisory services can help promote the best management practices of the frontiers to improve the TE, AE, and EE of the inefficient DMUs.

The present study seeks to investigate efficiency of roads in provinces in Iran. Considering the resource limitation, prioritisation of roads seems necessary for better allocation of resources. Since it is possible to investigate decision-making units (DMUs) where there are several inputs and outputs without considering any weight for them in data envelopment analysis, this method is very suitable for ranking. In order to analyse efficiency, BCC model with variable return to scale was used and given that a number of DMUs become efficient after solving the model, Anderson-Peterson method was employed to determine strong efficiency. Considering that one of the advantages of data envelopment analysis is the possibility of offering a progress model for inefficient DMUs, the essential changes in inputs and outputs of inefficient DMUs for being efficient have been mentioned.

Survey on rank preservation and rank reversal in data envelopment analysis

Data envelopment analysis (DEA) is a tool for measuring the relative efficiency of decision-making units (DMUs) that perform similar tasks. One of the uses of the DEA is the establishment of benchmarks for inefficient DMUs. These benchmarks play the role of target for inefficient DMUs. However, traditional DEA (TDEA) models fail to rank the efficient DMUs, and as a result, they cannot determine a benchmark for the inefficient DMUs. To overcome this shortcoming, this paper proposes virtual network DEA (NDEA) approach. The proposed method not only ranks the whole inefficient and efficient DMUs but also establishes a virtual benchmark called ideal DMU. The case study demonstrates the applicability of the proposed model.

Globally, insurance and takaful companies play a crucial role in economic and financial development of a country. Malaysia’s insurance sector has undergone substantial changes compared to two decades ago. Due to intense competition observed in Malaysian insurance industry, efficiency measurement and a complete efficiency ranking of insurance companies are very important for the decision makers so that necessary changes and improvement can be made. Data Envelopment Analysis (DEA) is a non-parametric method that has been acknowledged as an effective method to measure the efficiency of homogeneous decision-making units (DMUs). The main advantage of DEA is its ability to handle multiple inputs and outputs. However, standard DEA has poor discrimination power since it generates too many efficient units especially when the number of DMUs under study is insufficient in comparison to number of inputs and outputs. It cannot discriminate efficient units and therefore unable to give a complete ranking of DMUs. This study aimed to overcome the ranking problem found in standard DEA by integrating DEA and Preference Ranking Organization Method for Enrichment Evaluations (PROMETHEE) II. The hybrid method was applied to evaluate the efficiency and ranking of 22 life insurance and takaful companies in Malaysia from the period of 2017 to 2018. Input variables used in this study comprised of fees and commission, and management expenses. Meanwhile, output variables were net premium and generated investment income. The proposed method involved two stages. In the first stage, DEA was applied to obtain the efficient scores for the DMUs. In the second stage, PROMETHEE II was implemented to rank the efficient units. This hybrid method has successfully obtained full ranking of all the insurance companies under study. The inefficient companies can learn strategies and practices from efficient companies to enhance their services. It is recommended for future research to integrate DEA with other Multi Criteria Decision Making (MCDM) methods such as TOPSIS and VIKOR to fully rank the DMUs.

Data envelopment analysis (DEA) allows evaluation of the relative efficiencies of similar entities, known as decision making units (DMUs), which consume the same types of resources and offer similar types of products. It is known that under certain circumstances, when the number of DMUs does not meet the DEA Golden Rule, that is, this number is not sufficiently large compared to the total number of inputs and outputs, traditional DEA models often yield solutions that identify too many DMUs as efficient. In fact, this weak discrimination power and unrealistic weight distribution presented by DEA models remain a major challenge, leading to the development of models to improve this performance, such as: multiple criteria data envelopment analysis (MCDEA), bi-objective multiple criteria data envelopment analysis, goal programming approaches to solve weighted goal programming (WGP–MCDEA) and extended–MCDEA. This paper proposes a new MCDEA model which is based on goal programming, with and without super efficiency concepts, and presents test results that show its advantages over the above cited models. A set of problems from the literature and real-world applications are used in these tests. The results show that the new MCDEA model provides better discrimination of DMUs in all tested problems, and provides a weight dispersion that is statistically equal to that obtained by other MCDEA models. An additional feature of the proposed model is that it allows the identification of the input and output variables that are most important to the problem, to make it easier for the decision maker to improve the efficiency of the DMUs involved. This is very useful in practice, because in general, the available resources are scarce, so it is a further advantage of the proposed MCDEA model over the others tested.

In recent years, social network analysis (SNA) has been combined with data envelopment analysis (DEA) to select benchmarks and fully rank decision-making-units (DMUs). However, such methods fail to identify the most suitable benchmarks for the group. On one hand, they may incorrectly identify an efficient DMU as a suitable benchmark for the group. On the other hand, they fail to recognise any inefficient DMU as a benchmark for the group. To address such limitations occurred in the conventional DEA-based SNA approaches, this study proposes a modified DEA–SNA method. The new approach selects preferred benchmarks from the perspective of the entire DMU group considering both efficient DMUs and inefficient DMUs as candidates. The proposed method can identify efficient DMUs that obtain little endorsement from inefficient DMUs and inefficient DMUs that could be good benchmarks because they are endorsed by many DMUs with worse performance. The benchmarking information helps decision-makers identify suitable benchmarks, especially for those units whose efficiency scores are too low to learn from the industry leaders. Two examples are used to illustrate the practicability and superiority of our proposed method.

Efficiency evaluation based on data envelopment analysis in the big data context

Data envelopment analysis (DEA) identifies an empirical efficient frontier of a set of peer decision making units (DMUs) with multiple inputs and outputs. The efficient frontier is characterized by the DMUs with an unity efficiency score. The performance of inefficient DMUs is characterized with respect to the identified efficient frontier. If the performance of inefficient DMUs deteriorates or improves (up to the frontier), the efficient DMUs still have an unity efficiency score. However, the performance of DMUs may be influenced by the context — e.g. a product may appear attractive against a background of less attractive alternatives and unattractive when compared to more attractive alternatives. With an application to Tokyo public libraries, the current paper presents and demonstrates a context-dependent DEA which measures the relative attractiveness of libraries on a specific performance level against libraries exhibiting poorer performance. The set of libraries are grouped into different levels of efficient frontiers. Each efficient frontier (on a specific performance level) is then used as evaluation context for the relative attractiveness. The performance of the efficient libraries changes as the inefficient libraries change their performance. The context-dependent DEA can also be used to differentiate the performance of efficient DMUs. The context-dependent DEA provides finer DEA results with respect to the performance of all DMUs.

The way to deal with flexible data from their stochastic presence point of view as output or input in the evaluation of efficiency of the decision-making units (DMUs) motivates new perspectives in modeling and solving data envelopment analysis (DEA) in the presence of flexible variables. Because the orientation of flexible data is not pre-determined, and because the number of DMUs is fixed and all the DMUs are independent, flexible data can be treated as random variable in terms of both input and output selection. As a result, the selection of flexible variable as input or output for n DMUs can be regarded as binary random variable. Assuming the randomness of choosing flexible data as input or output, we deal with DEA models in the presence of flexible data whose input or output orientation determines a binomial distribution function. This study provides a new insight to classify flexible variable and investigates the input or output status of a variable using a stochastic model. The proposed model obviates the problems caused by the use of the large M number and using its different values in previous models. In addition, it can obtain the most appropriate efficiency value for decision-making units by assigning the chance of choosing the orientation of flexible variable to the model itself. The proposed method is compared with other available methods by employing numerical and empirical examples.

Data envelopment analysis (DEA) is a non-parametric technique for measuring the efficiency of decision making units (DMU) with exact input and output terms. So far, a number of DEA models with interval data have been developed. The radial DEA model such as CCR or BCC model with interval data are well known as basic DEA models with interval data. However, these radial DEA models have some defects. In this study, we develop a new approach based upon the slacks-based measurement (SBM) of super-efficiency model for dealing with interval data in DEA. The SBM super-efficiency DEA model can not only discriminate the performance of efficient DMUs from inefficient ones, but can distinguish between the efficient DMUs. In addition, this model can overcome the drawback that the super-efficiency model can be infeasible under variable returns to scale (VRS) assumption. An illustrative example is provided to verify the idea of this paper and show its potential application and validity.

Data envelopment analysis (DEA) with considering the best condition for each decision making unit (DMU) assesses the relative efficiency for it and divides a homogenous group of DMUs into two categories: efficient and inefficient, but traditional DEA models can not rank efficient DMUs. Although some models were introduced for ranking efficient DMUs, Franklin Lio & Hsuan peng (2008), proposed a common weights analysis (CWA) approach for ranking them. These DMUs are ranked according to the efficiency score weighted by the common set of weights and shadow prices. This study shows there are some cases that shadow prices of efficient DMUs are equal, hence this method is not applicable for ranking them. Next, we propose a new method for ranking units with equal shadow prices.

When the number of Decision-Making Units (DMUs) is not large enough compared to the total number of input parameters and outputs, traditional Data Envelopment Analysis (DEA) and Network Data Envelopment Analysis (NDEA) models often produce solutions that identify many DMUs as efficient, in addition to obtaining unrealistic weight distributions. In fact, this poor discrimination power and unrealistic weight distribution presented by DEA and NDEA models remain a major challenge, leading to the development of models to improve this performance. Thus, this paper proposes a Multiple Criteria Network Data Envelopment Analysis (MCNDEA), based on relational NDEA models. The idea of this model is to be used in network structures. To test the MCNDEA model, one a real instance linked to an evaluation problem of academic departments of a public university was used. Other instances were also used to validate the proposed MCNDEA, and these tests are included in the supplementary files. Finally, it should be noted that, in summary, the model proposed in this article had greater discrimination power of the analyzed DMUs, being able to identify the most efficient departments in each of the considered stages, besides pointing out to the University Management points of improvement regarding the best use of its resources for each department.

Background:The Integrated Child Development Services (ICDS) scheme was launched in 1975 for the improvement in maternal and child health and nutrition. The services under ICDS are implemented through Anganwadi centers (AWCs) and are delivered by Anganwadi workers (AWWs) at grassroots level. To evaluate the efficiency of all AWCs located in the field practice area of the medical college in North India, using data envelopment analysis (DEA) technique.Materials and Methods:A cross-sectional study was conducted in the catchment area of rural health training center. Each of the 15 AWCs was considered a decision-making unit (DMU), and physical structure and logistics were included as input variables; and percentage of beneficiaries receiving supplementary nutrition and health checkups were included as output variables. DEA technique was used to calculate the efficiency score for each DMU with the help of DEAOS free online software.Results:DMU 14th was found to be most efficient (100%) and DMU second was the least efficient (45%). DMU 13 and 14 demonstrated a level of performance that is superior to all other DMUs. DMU 13th and 14th were, therefore, considered 100% efficient. DEA analysis showed that total output increases and/or input reductions were required for making the inefficient DMUs efficient.Conclusion:Most of the AWCs were inefficient and an improvement in the infrastructure and logistics is likely to improve the efficiency of AWCs.

Data envelopment analysis (DEA) facilitates relative performance estimation of homogenous decision-making units (DMUs), whereas dynamic DEA pertains to intertemporal elements like carryovers for measuring DMUs’ efficiencies over time. To enhance DMU’s performance, there is a need to measure its efficiency with dynamic structure and further to determine input-output target points for making significant improvements in an inefficient DMU. Therefore, the present study proposes a relational dynamic DEA approach that comprises a dynamic structure in which periods are connected through good and bad links/carry-overs along with desirable and undesirable outputs and utilizes interval data. The system and period efficiency intervals are derived using a unique set of weights based on common weights methodology. Moreover, the relationship between the complement of the lower (upper) bound system and period efficiencies is established. Lastly, the input-output target points are suggested to improve DMUs in terms of upper-bound system efficiencies in a dynamic environment. To validate the usefulness of the proposed approach, a case study in Indian banks for the period 2017– 2021 is presented. This study investigates the impact of stressed assets (bad link/carryover) and loss due to non-performing assets (bad/undesirable output) on banks’ dynamic efficiency. The results are compared with the static structure. The findings, targets, and implications of the study can assist bank experts and policymakers in formulating policies/strategies for further improvements.