Scholarlink

<p>This study introduces a cognition-enhanced framework for geospatial decision-making by integrating Fuzzy Formal Concept Analysis (FCA), the Surprisingly Popular (SP) method, and a Large Language Model (GPT-4o). Our approach captures cognitive influences that are often overlooked in traditional geospatial analyses. Fuzzy FCA is used to extract interpretable concept hierarchies from spatial data, while the GPT-4o model estimates SP scores, identifying choices that reflect underlying cognitive biases. These cognitively informed features are combined within machine learning models, improving both prediction accuracy and interpretability. Experiments on real-world urban mobility and environmental risk scenarios demonstrate significant performance gains, with models like XGBoost achieving an accuracy of 0.8412. We also introduce a novel method for evaluating the cognitive validity of LLM-generated model explanations, which involves assessing how well these explanations align with human intuition and reasoning. Our results show that incorporating cognitive elements into geospatial models not only improves outcomes but also bridges the gap between data-driven predictions and human decision-making. This framework offers broad potential for applications in GIS, urban planning, and environmental management.</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

Cognition-enhanced geospatial decision framework integrating fuzzy FCA, surprisingly popular method, and a large language model

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月2日

查看全文
  • Bongjae Kwon,
  • Kiyun Yu

摘要

This study introduces a cognition-enhanced framework for geospatial decision-making by integrating Fuzzy Formal Concept Analysis (FCA), the Surprisingly Popular (SP) method, and a Large Language Model (GPT-4o). Our approach captures cognitive influences that are often overlooked in traditional geospatial analyses. Fuzzy FCA is used to extract interpretable concept hierarchies from spatial data, while the GPT-4o model estimates SP scores, identifying choices that reflect underlying cognitive biases. These cognitively informed features are combined within machine learning models, improving both prediction accuracy and interpretability. Experiments on real-world urban mobility and environmental risk scenarios demonstrate significant performance gains, with models like XGBoost achieving an accuracy of 0.8412. We also introduce a novel method for evaluating the cognitive validity of LLM-generated model explanations, which involves assessing how well these explanations align with human intuition and reasoning. Our results show that incorporating cognitive elements into geospatial models not only improves outcomes but also bridges the gap between data-driven predictions and human decision-making. This framework offers broad potential for applications in GIS, urban planning, and environmental management.