Using Neural Networks to Determine Sugeno Measures by Statistics

Area of Science:

• Decision Science
• Artificial Intelligence
• Fuzzy Logic

Background:

• Traditional weighted average methods struggle with multi-attribute synthetic evaluation.
• Determining fuzzy measures for Choquet and Sugeno integrals is complex due to subjectivity and non-additivity.
• Existing methods face challenges in accurately capturing attribute interactions.

Purpose of the Study:

• To develop an optimized method for determining Sugeno measures using Choquet integrals.
• To address the difficulties in defining fuzzy measures for synthetic evaluation.
• To leverage neural networks for inverse problems in multi-attribute analysis.

Main Methods:

• Utilized Choquet integrals and Sugeno integrals for synthetic evaluation.
• Employed a neural network algorithm to optimize the inverse problem.
• Applied statistical analysis to given data for measure determination.

Main Results:

• Successfully determined Sugeno measures through an optimized inverse problem approach.
• Demonstrated the capability of neural networks in fuzzy measure optimization.
• Showcased the Choquet integral as a generalization of weighted averages.

Conclusions:

• The proposed neural network method offers an effective solution for determining fuzzy measures.
• This approach enhances synthetic evaluation accuracy in multi-attribute scenarios.
• The methodology has broad applicability in multivariate analysis, decision making, and pattern recognition.