Scholarlink

<p>Evaluating service quality in the automotive after-sales industry demands precise measurement and actionable interpretability to facilitate targeted improvements. To address the complexity of service systems and the inherent uncertainty of business data in this domain, this paper presents an innovative evaluation framework integrating multi-classifier systems with causal reasoning. Leveraging the interval-valued Fermatean fuzzy set, this paper systematically models the bidirectional uncertainty inherent in after-sales service data. Two novel dynamic classifier selection indices are developed to optimize ensemble performance. By incorporating counterfactual analysis with attribute weighting mechanisms, the proposed model generates interpretable service quality ratings accompanied by causal explanations. Experimental results show the model outperforms five baselines with 7/10 datasets achieving higher accuracy (average +3.70%), 8/10 optimal F1-scores (average +3.40%), and an average 93.07% accuracy in real-world scenarios. Notably, it achieves an average AUC of 0.915, surpassing all compared models. For interpretability, the framework is validated against SHAP, FICO, and LIME through feature distribution analysis, Jaccard similarity assessment, and top-1 explanation intersection size distribution, demonstrating superior transparent decision support alongside high predictive performance.</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

Model for service quality evaluation in multi-value chain networks using causality and multiple classifier system

  • 期刊论文

  • 发表日期：

    2026年3月16日

查看全文
  • Jingxiong Qiu,
  • Linfu Sun

摘要

Evaluating service quality in the automotive after-sales industry demands precise measurement and actionable interpretability to facilitate targeted improvements. To address the complexity of service systems and the inherent uncertainty of business data in this domain, this paper presents an innovative evaluation framework integrating multi-classifier systems with causal reasoning. Leveraging the interval-valued Fermatean fuzzy set, this paper systematically models the bidirectional uncertainty inherent in after-sales service data. Two novel dynamic classifier selection indices are developed to optimize ensemble performance. By incorporating counterfactual analysis with attribute weighting mechanisms, the proposed model generates interpretable service quality ratings accompanied by causal explanations. Experimental results show the model outperforms five baselines with 7/10 datasets achieving higher accuracy (average +3.70%), 8/10 optimal F1-scores (average +3.40%), and an average 93.07% accuracy in real-world scenarios. Notably, it achieves an average AUC of 0.915, surpassing all compared models. For interpretability, the framework is validated against SHAP, FICO, and LIME through feature distribution analysis, Jaccard similarity assessment, and top-1 explanation intersection size distribution, demonstrating superior transparent decision support alongside high predictive performance.