CRITID: enhancing CRITIC with advanced independence testing for robust multi-criteria decision-making

Area of Science:

• Decision Science
• Data Science
• Statistical Modeling

Background:

• Determining criteria weights is essential for multi-criteria decision-making and model evaluation.
• The big data era necessitates advanced methods for complex problem analysis.
• Traditional CRiteria Importance Through Intercriteria Correlation (CRITIC) methods using Pearson correlation may fail with nonlinear data.

Purpose of the Study:

• To refine the CRITIC method for better accommodation of nonlinear relationships.
• To enhance the robustness of criteria weight determination in complex datasets.
• To introduce a novel method for assessing intercriteria dependence.

Main Methods:

• Development of the CRiteria Importance Through Intercriteria Dependence (CRITID) method.
• Utilization of advanced independence testing, including distance correlation.
• Application and comparison across diverse data distributions.

Main Results:

• The CRITID method demonstrates enhanced rationality and robustness compared to traditional CRITIC.
• Improved assessment of intercriteria relationships, particularly nonlinear ones.
• Validation across varied data distributions confirms method efficacy.

Conclusions:

• CRITID offers a more accurate and dependable framework for multi-criteria decision-making.
• The novel method significantly benefits model evaluation in the context of big data.
• Enhanced handling of nonlinearities provides a more robust approach to complex data analysis.