Classification of soft decision-making methods via fuzzy parameterized fuzzy soft matrices and their...

Area of Science:

• Machine Learning
• Artificial Intelligence
• Data Science

Background:

• Fuzzy parameterized fuzzy soft matrices (fpfs-matrices) provide a robust framework for uncertainty modeling.
• A gap exists in large-scale comparative studies of soft decision-making (SDM) algorithms derived from fpfs-matrices within machine learning.

Purpose of the Study:

• To comprehensively evaluate and classify 35 SDM algorithms based on fpfs-matrices.
• To benchmark these algorithms on diverse datasets using standard classification metrics.
• To identify top-performing algorithms and provide practical guidance for selecting SDM methods.

Main Methods:

• Utilized the Comparison Matrix-Based Fuzzy Parameterized Fuzzy Soft Classifier (FPFS-CMC).
• Evaluated 35 SDM algorithms across ten UCI Machine Learning Repository datasets.
• Assessed performance using accuracy, precision, recall, specificity, and F1-score, with statistical validation via Friedman and Nemenyi tests.

Main Results:

• SDM methods employing fpfs-matrices show competitive classification performance in uncertain environments.
• Identified top-performing algorithms: A19 (Rank 1), YHX14 (Rank 2), and a tie for Rank 3 (VMH16, AKO18o, A19/2).
• Statistical tests confirmed the significance of performance differences.

Conclusions:

• SDM methods based on fpfs-matrices are effective for machine learning classification tasks with uncertainty.
• The study offers a decision-support framework and practical insights for algorithm selection.
• This research serves as a theoretical reference for future work in uncertainty-based decision-making.