A multi-dimensional decision framework based on the XGBoost algorithm and the constrained parametric approach.

Area of Science:

• Decision Science
• Data Analytics
• Machine Learning

Background:

• Organizations face challenges in decision-making due to data complexity and uncertainty.
• Need for interpretable and data-driven analytical methods is increasing.
• Existing methods may not adequately handle both quantitative data and qualitative linguistic variables.

Purpose of the Study:

• To propose a multi-dimensional decision framework integrating quantitative data and qualitative variables.
• To develop a method for calculating attribute weights using historical data and machine learning.
• To provide a scientific decision matrix for rational decision-making in uncertain environments.

Main Methods:

• Utilized nested probabilistic linguistic term sets for decision modeling.
• Applied the XGBoost algorithm to determine attribute importance and weights from historical data.
• Employed a constrained parametric approach to establish membership functions for linguistic variables.

Main Results:

• Developed a flexible and stable decision framework.
• Successfully generated a scientific decision matrix by integrating data and linguistic variables.
• Demonstrated the framework's efficacy through a bank credit ranking case study.

Conclusions:

• The proposed framework offers a robust approach for capturing, integrating, and analyzing data and linguistic variables.
• It supports managers and policymakers in determining attribute weights and selecting appropriate decision methods.
• The framework enhances both theoretical understanding and practical application of uncertain and cognitive decision-making.