A data-driven decision support framework for DEA target setting: an explainable AI approach

Mustafa Jahangoshai Rezaee; Mohsen Abbaspour Onari; Morteza Saberi

doi:10.1016/j.engappai.2023.107222

A data-driven decision support framework for DEA target setting: an explainable AI approach

Mustafa Jahangoshai Rezaee (Corresponding author)
, Mohsen Abbaspour Onari (Corresponding author)
, Morteza Saberi (Corresponding author)

Research output: Contribution to journal › Article › Academic › peer-review

13 Citations (Scopus)

130 Downloads (Pure)

Abstract

The intention of target setting for Decision-Making Units (DMUs) in Data Envelopment Analysis (DEA) is to perform better than their peers or reach a reference efficiency level. However, most of the time, the logic behind the target setting is based on mathematical models, which are not achievable in practice. Besides, these models are based on decreasing/increasing inputs/outputs that might not be feasible based on DMU's potential in the real world. We propose a data-driven decision support framework to set actionable and feasible targets based on vital inputs-outputs for target setting. To do so, DMUs are classified in their corresponding Efficiency Frontier (EF) levels based on multiple EFs approach and a machine learning classifier. Then, the vital inputs-outputs are determined using an Explainable Artificial Intelligence (XAI) method. Finally, a Multi-Objective Counterfactual Explanation is developed based on DEA (MOCE-DEA) to lead DMU in reaching the reference EF by adjusting actionable and feasible inputs-outputs. We studied Iranian hospitals to evaluate the proposed framework and presented two cases to demonstrate its mechanism. The results show that the performance of the DMUs is improved to reach the reference EF for studied cases. Then, a validation was conducted with the primal DEA model to show the robust improvement of DMUs after adjusting their original value based on the generated solutions by the proposed framework. It demonstrates that the adjusted values can also improve DMUs' performance in the primal DEA model.

Original language	English
Article number	107222
Number of pages	14
Journal	Engineering Applications of Artificial Intelligence
Volume	127
Issue number	part A
DOIs	https://doi.org/10.1016/j.engappai.2023.107222
Publication status	Published - Jan 2024

Bibliographical note

Publisher Copyright:
© 2023 The Author(s)

Funding

The main aim of the current research is to propose a data-driven decision support framework according to the XAI approach for DEA benchmarking. The proposed framework first creates multiple EFs by the primal CCR model with their corresponding labels to create multiple EFs to propose a hierarchical strategy for target setting. According to the RF classifier, the labeled dataset is classified, and the prediction's outcome is utilized to obtain the dominant features of DMUs. Then, it is possible to set the actionable and feasible target for a DMU by finding its equivalent peer in the reference EF by Euclidean distance. LIME can extract dominant features for them that positively contribute to the prediction's outcome. Afterward, MOCE-DEA is implemented for target setting by helping DMUo to improve its PS to reach the ES of DMUp by adjusting actionable and feasible features. For this purpose, DMUo is considered a new member of the reference EF. Two objectives should be met: i) the distance of actionable and feasible features of DMUo with DMUp should be minimized; ii) the ES of DMUo should be maximized. MOPSO solves this model, and the best solution with the highest ES among Pareto optimal solutions for each simulation batch is collected. For evaluating the model's performance and obtaining a vivid insight into the target setting, applied adjustments based on MOCE-DEA on DMUo are verified by the primal CCR model. The robust evaluation shows that based on optimized values DMUo can reach the reference EF in the primal CCR model as well. The successful performance of this approach makes it possible to set realistic targets. Besides, it can support DMs in finding feasible targets, and rest assured that target setting is based on vital inputs-outputs that are adjustable.

Keywords

Benchmarking
Data envelopment analysis
Explainable artificial intelligence
LIME
Multi-objective counterfactual explanation
Target setting

Access to Document

10.1016/j.engappai.2023.107222

pdfFinal published version, 3.3 MBLicence: CC BY

Cite this

@article{70604887b55b4cb19886296f2d3374ba,

title = "A data-driven decision support framework for DEA target setting: an explainable AI approach",

abstract = "The intention of target setting for Decision-Making Units (DMUs) in Data Envelopment Analysis (DEA) is to perform better than their peers or reach a reference efficiency level. However, most of the time, the logic behind the target setting is based on mathematical models, which are not achievable in practice. Besides, these models are based on decreasing/increasing inputs/outputs that might not be feasible based on DMU's potential in the real world. We propose a data-driven decision support framework to set actionable and feasible targets based on vital inputs-outputs for target setting. To do so, DMUs are classified in their corresponding Efficiency Frontier (EF) levels based on multiple EFs approach and a machine learning classifier. Then, the vital inputs-outputs are determined using an Explainable Artificial Intelligence (XAI) method. Finally, a Multi-Objective Counterfactual Explanation is developed based on DEA (MOCE-DEA) to lead DMU in reaching the reference EF by adjusting actionable and feasible inputs-outputs. We studied Iranian hospitals to evaluate the proposed framework and presented two cases to demonstrate its mechanism. The results show that the performance of the DMUs is improved to reach the reference EF for studied cases. Then, a validation was conducted with the primal DEA model to show the robust improvement of DMUs after adjusting their original value based on the generated solutions by the proposed framework. It demonstrates that the adjusted values can also improve DMUs' performance in the primal DEA model.",

keywords = "Benchmarking, Data envelopment analysis, Explainable artificial intelligence, LIME, Multi-objective counterfactual explanation, Target setting",

author = "\{Jahangoshai Rezaee\}, Mustafa and \{Abbaspour Onari\}, Mohsen and Morteza Saberi",

note = "Publisher Copyright: {\textcopyright} 2023 The Author(s)",

year = "2024",

month = jan,

doi = "10.1016/j.engappai.2023.107222",

language = "English",

volume = "127",

journal = "Engineering Applications of Artificial Intelligence",

issn = "0952-1976",

publisher = "Elsevier",

number = "part A",

}

A data-driven decision support framework for DEA target setting: an explainable AI approach. / Jahangoshai Rezaee, Mustafa (Corresponding author); Abbaspour Onari, Mohsen (Corresponding author); Saberi, Morteza (Corresponding author).
In: Engineering Applications of Artificial Intelligence, Vol. 127, No. part A, 107222, 01.2024.

Research output: Contribution to journal › Article › Academic › peer-review

TY - JOUR

T1 - A data-driven decision support framework for DEA target setting

T2 - an explainable AI approach

AU - Jahangoshai Rezaee, Mustafa

AU - Abbaspour Onari, Mohsen

AU - Saberi, Morteza

PY - 2024/1

Y1 - 2024/1

N2 - The intention of target setting for Decision-Making Units (DMUs) in Data Envelopment Analysis (DEA) is to perform better than their peers or reach a reference efficiency level. However, most of the time, the logic behind the target setting is based on mathematical models, which are not achievable in practice. Besides, these models are based on decreasing/increasing inputs/outputs that might not be feasible based on DMU's potential in the real world. We propose a data-driven decision support framework to set actionable and feasible targets based on vital inputs-outputs for target setting. To do so, DMUs are classified in their corresponding Efficiency Frontier (EF) levels based on multiple EFs approach and a machine learning classifier. Then, the vital inputs-outputs are determined using an Explainable Artificial Intelligence (XAI) method. Finally, a Multi-Objective Counterfactual Explanation is developed based on DEA (MOCE-DEA) to lead DMU in reaching the reference EF by adjusting actionable and feasible inputs-outputs. We studied Iranian hospitals to evaluate the proposed framework and presented two cases to demonstrate its mechanism. The results show that the performance of the DMUs is improved to reach the reference EF for studied cases. Then, a validation was conducted with the primal DEA model to show the robust improvement of DMUs after adjusting their original value based on the generated solutions by the proposed framework. It demonstrates that the adjusted values can also improve DMUs' performance in the primal DEA model.

AB - The intention of target setting for Decision-Making Units (DMUs) in Data Envelopment Analysis (DEA) is to perform better than their peers or reach a reference efficiency level. However, most of the time, the logic behind the target setting is based on mathematical models, which are not achievable in practice. Besides, these models are based on decreasing/increasing inputs/outputs that might not be feasible based on DMU's potential in the real world. We propose a data-driven decision support framework to set actionable and feasible targets based on vital inputs-outputs for target setting. To do so, DMUs are classified in their corresponding Efficiency Frontier (EF) levels based on multiple EFs approach and a machine learning classifier. Then, the vital inputs-outputs are determined using an Explainable Artificial Intelligence (XAI) method. Finally, a Multi-Objective Counterfactual Explanation is developed based on DEA (MOCE-DEA) to lead DMU in reaching the reference EF by adjusting actionable and feasible inputs-outputs. We studied Iranian hospitals to evaluate the proposed framework and presented two cases to demonstrate its mechanism. The results show that the performance of the DMUs is improved to reach the reference EF for studied cases. Then, a validation was conducted with the primal DEA model to show the robust improvement of DMUs after adjusting their original value based on the generated solutions by the proposed framework. It demonstrates that the adjusted values can also improve DMUs' performance in the primal DEA model.

KW - Benchmarking

KW - Data envelopment analysis

KW - Explainable artificial intelligence

KW - LIME

KW - Multi-objective counterfactual explanation

KW - Target setting

UR - https://www.scopus.com/pages/publications/85173389912

U2 - 10.1016/j.engappai.2023.107222

DO - 10.1016/j.engappai.2023.107222

M3 - Article

AN - SCOPUS:85173389912

SN - 0952-1976

VL - 127

JO - Engineering Applications of Artificial Intelligence

JF - Engineering Applications of Artificial Intelligence

IS - part A

M1 - 107222

ER -

A data-driven decision support framework for DEA target setting: an explainable AI approach

Abstract

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this