Scholarlink

<p>Environmental sustainability is a 21st-century challenge that requires advanced computational frameworks to evaluate ecological design. Existing systems typically employ single-dimensional measures, overlooking the multimodal and spatiotemporal complexity of sustainability indicators. A 19-year evaluation framework including 54 sustainability metrics from 173 nations is proposed in this research. The goals are to develop a multimodal graph-based representation of environmental, social, and economic variables and to create an intelligent system that provides holistic and interpretable evaluations. The Ecological Graph-based Multimodal Assessment Engine (EcoGraph-MMAE) is a data-driven framework for continuous prediction of the Environmental Sustainability Index (ESI) for environmental design evaluation. The methodology employs a spatiotemporal graph attention algorithm that combines tabular indicators, international organizations’ policy indicators, and geospatial adjacency. Graph Attention Networks (GATs) and temporal modeling (LSTM and 3D-CNNs) enable the system to learn regional dependencies and long-term sustainability patterns. The results show that multimodal fusion improves predictive accuracy by 15–18% over unimodal baselines. The approach reduces the RMSE by 12% and improves the F1-score by 0.14, demonstrating broad generalization across various areas. Results show that climate action (SDG 13), inequality reduction (SDG 10), and sustainable energy (SDG 7) indicators have the most significant impact on environmental design performance. The model also provides SDG-aligned country-level evaluation ratings. Ultimately, EcoGraph-MMAE employs multimodal graph neural networks to assess ecological design in a scalable and data-driven manner.</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 comprehensive evaluation system for environmental design supported by multimodal graph neural networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月13日

查看全文
  • Jing Hu,
  • Lin Li

摘要

Environmental sustainability is a 21st-century challenge that requires advanced computational frameworks to evaluate ecological design. Existing systems typically employ single-dimensional measures, overlooking the multimodal and spatiotemporal complexity of sustainability indicators. A 19-year evaluation framework including 54 sustainability metrics from 173 nations is proposed in this research. The goals are to develop a multimodal graph-based representation of environmental, social, and economic variables and to create an intelligent system that provides holistic and interpretable evaluations. The Ecological Graph-based Multimodal Assessment Engine (EcoGraph-MMAE) is a data-driven framework for continuous prediction of the Environmental Sustainability Index (ESI) for environmental design evaluation. The methodology employs a spatiotemporal graph attention algorithm that combines tabular indicators, international organizations’ policy indicators, and geospatial adjacency. Graph Attention Networks (GATs) and temporal modeling (LSTM and 3D-CNNs) enable the system to learn regional dependencies and long-term sustainability patterns. The results show that multimodal fusion improves predictive accuracy by 15–18% over unimodal baselines. The approach reduces the RMSE by 12% and improves the F1-score by 0.14, demonstrating broad generalization across various areas. Results show that climate action (SDG 13), inequality reduction (SDG 10), and sustainable energy (SDG 7) indicators have the most significant impact on environmental design performance. The model also provides SDG-aligned country-level evaluation ratings. Ultimately, EcoGraph-MMAE employs multimodal graph neural networks to assess ecological design in a scalable and data-driven manner.