Fraud Detection Dataset Evaluation Using Complex Hesitant Fuzzy Rough Multi-Attribute Decision Making Method

Fraud detection has become a critical research area due to the increasing volume of digital transactions and the complex, uncertain nature of fraudulent activities.Although several fraud detection datasets and machine learning models have been proposed, most existing approaches rely on crisp evaluations that fail to effectively handle uncertainty, hesitation, and ambiguity present in real-world fraud data.To address this limitation, this paper proposes a novel framework for evaluating fraud detection datasets using a Complex Hesitant Fuzzy Rough Multi-Attribute Decision Making (CHFR MADM) method.The proposed approach integrates complex hesitant fuzzy sets to capture expert hesitation and uncertainty, while rough set theory is employed to manage boundary regions and indiscernibility among datasets.Multiple datasets and performance attributes, including class imbalance, noise sensitivity, robustness, and detection accuracy, are considered as decision criteria.The CHFR MADM framework aggregates these criteria to rank fraud detection datasets in a more reliable and flexible manner compared to traditional crisp and single-metric evaluation techniques.Experimental analysis using benchmark fraud datasets demonstrates that the proposed method provides consistent and informative rankings under uncertain conditions.The results confirm that the CHFR MADM approach is effective in supporting decision-making in fraud detection research and can serve as a valuable tool for selecting suitable datasets in complex and uncertain environments.