Scholarlink

<p>The aviation industry is increasingly using deep learning and machine learning models to improve the prediction and prevention of security threats. Many studies have proposed advanced methods to detect prohibited items in X-ray security images. However, there is still no clear agreement on how these models should be evaluated and ranked when many performance criteria are considered together. Therefore, this paper presents an integrated Multi-criteria Decision Making (MCDM) framework. The framework constructs a decision matrix by crossing 12 models (combining InceptionV3 with twelve supervised machine learning classifiers) with 7 evaluation criteria. Entropy method is used to determine objective weights for the evaluation criteria. After that, TOPSIS, VIKOR, and EDAS are adopted to rank the Models based on calculated weight values. The Entropy results revealed that Precision is the most influential criterion with a weight value of 0.1483. The ranking results showed that the InceptionV3-Naive Bayes (M<sub>1</sub>) achieved the best performance according to TOPSIS with a <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\({c}_{i}\)</EquationSource> </InlineEquation> value of 0.9814 and also according to VIKOR with a <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(Q\)</EquationSource> </InlineEquation> value of 0.0132. In contrast, the InceptionV3–Logistic Regression (M<sub>7</sub>) obtained the highest rank using the EDAS method with a score of 0.9904. The variations in selecting the best ranked model back to the methods’ assumptions and characteristics. A sensitivity analysis was also conducted to test the stability of the rankings when the criterion weights change. The results indicated that VIKOR is more stable than TOPSIS and EDAS. The proposed framework can provide useful insights for the aviation industry, security administrators, researchers and practitioners.</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

A hybrid MCDM perspective on model selection for X-ray image classification: comparative insights from TOPSIS, VIKOR, and EDAS

  • 期刊论文

  • 发表日期：

    2026年5月14日

查看全文
  • Wafaa Ayoub Kassara,
  • Baydaa Flayyih Hasan,
  • Bushra Raad Zahi

摘要

The aviation industry is increasingly using deep learning and machine learning models to improve the prediction and prevention of security threats. Many studies have proposed advanced methods to detect prohibited items in X-ray security images. However, there is still no clear agreement on how these models should be evaluated and ranked when many performance criteria are considered together. Therefore, this paper presents an integrated Multi-criteria Decision Making (MCDM) framework. The framework constructs a decision matrix by crossing 12 models (combining InceptionV3 with twelve supervised machine learning classifiers) with 7 evaluation criteria. Entropy method is used to determine objective weights for the evaluation criteria. After that, TOPSIS, VIKOR, and EDAS are adopted to rank the Models based on calculated weight values. The Entropy results revealed that Precision is the most influential criterion with a weight value of 0.1483. The ranking results showed that the InceptionV3-Naive Bayes (M1) achieved the best performance according to TOPSIS with a \({c}_{i}\) value of 0.9814 and also according to VIKOR with a \(Q\) value of 0.0132. In contrast, the InceptionV3–Logistic Regression (M7) obtained the highest rank using the EDAS method with a score of 0.9904. The variations in selecting the best ranked model back to the methods’ assumptions and characteristics. A sensitivity analysis was also conducted to test the stability of the rankings when the criterion weights change. The results indicated that VIKOR is more stable than TOPSIS and EDAS. The proposed framework can provide useful insights for the aviation industry, security administrators, researchers and practitioners.