Scholarlink

<p>The growing reliance on artificial intelligence (AI) in high-security domains such as X-ray threat detection in the context of baggage screening has highlighted the need for robust, interpretable, and data-driven evaluation frameworks. Despite extensive efforts to enhance classification model performance, no existing study has ranked these models within the context of X-ray baggage imagery using multi-criteria decision-making (MCDM) techniques. This review addresses this gap by conducting a comprehensive analysis of classification models and their associated evaluation metrics, identifying limitations in metric selection practices—such as redundancy, lack of prioritization, and inconsistent evaluation frameworks. Drawing insights from related domains like medical imaging, the study critically examines existing MCDM weighting and ranking methods, benchmarking techniques, and validation procedures. The limitations of conventional methods, including TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution), VIKOR (VIekriterijumsko KOmpromisno Rangiranje), and EDAS (Evaluation based on Distance from Average Solution), are discussed in terms of distance measures, sensitivity to uncertainty, and inability to integrate subjective and objective weights. To address these issues, the paper proposes a novel hybrid framework that integrates Grey Relational Coefficient (GRC)-based ranking methods (GRC, GRC–TOPSIS, GRC–VIKOR, GRC–EDAS) with a fuzzy-enhanced, entropy-supported Interval-valued Spherical Fuzzy Sets–Full Consistency Method (IvSFS–FUCOM) weighting strategy. This approach offers a more comprehensive, uncertainty-aware solution for evaluating classification models. Furthermore, the review highlights the role of dataset generation, benchmarking checklists, sensitivity-correlation analyses, and systematic ranking tests as critical tools for validating MCDM frameworks. The recommended methodology not only consolidates existing evaluation practices but also highlights its novelty by bridging theoretical gaps in MCDM-based ranking, validating results with multiple verification tools, and offering a reliable decision-support system adaptable to security screening and other high-stakes domains.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Ensemble-based decision support for X-ray threat detection: a review and hybrid MCDM framework under uncertainty

  • 期刊论文

  • 发表日期：

    2026年5月5日

查看全文
  • Esam Motashar Aday Almahdi,
  • Soong Der Chen,
  • Mohd Hazli Mohamed Zabil

摘要

The growing reliance on artificial intelligence (AI) in high-security domains such as X-ray threat detection in the context of baggage screening has highlighted the need for robust, interpretable, and data-driven evaluation frameworks. Despite extensive efforts to enhance classification model performance, no existing study has ranked these models within the context of X-ray baggage imagery using multi-criteria decision-making (MCDM) techniques. This review addresses this gap by conducting a comprehensive analysis of classification models and their associated evaluation metrics, identifying limitations in metric selection practices—such as redundancy, lack of prioritization, and inconsistent evaluation frameworks. Drawing insights from related domains like medical imaging, the study critically examines existing MCDM weighting and ranking methods, benchmarking techniques, and validation procedures. The limitations of conventional methods, including TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution), VIKOR (VIekriterijumsko KOmpromisno Rangiranje), and EDAS (Evaluation based on Distance from Average Solution), are discussed in terms of distance measures, sensitivity to uncertainty, and inability to integrate subjective and objective weights. To address these issues, the paper proposes a novel hybrid framework that integrates Grey Relational Coefficient (GRC)-based ranking methods (GRC, GRC–TOPSIS, GRC–VIKOR, GRC–EDAS) with a fuzzy-enhanced, entropy-supported Interval-valued Spherical Fuzzy Sets–Full Consistency Method (IvSFS–FUCOM) weighting strategy. This approach offers a more comprehensive, uncertainty-aware solution for evaluating classification models. Furthermore, the review highlights the role of dataset generation, benchmarking checklists, sensitivity-correlation analyses, and systematic ranking tests as critical tools for validating MCDM frameworks. The recommended methodology not only consolidates existing evaluation practices but also highlights its novelty by bridging theoretical gaps in MCDM-based ranking, validating results with multiple verification tools, and offering a reliable decision-support system adaptable to security screening and other high-stakes domains.