Scholarlink

<p>Conflict is ubiquitous and unavoidable in human life and society. Moreover, predicting chaotic conflicts with sensitivity to the initial conditions generated by a fuzzy conflict model is challenging. This paper introduces a novel chaotic status prediction method for the fuzzy conflict model influenced by external forces, applying a physics-informed neural network (PINN), which integrates physical laws into machine learning, to address the challenges posed by limited data availability. The model is trained with only 20%, 40%, and 60% of the total data, and its performance is evaluated using the mean squared error, symmetric mean absolute percentage error, and <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(R^{\textrm{2}}\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>R</mi> <mtext>2</mtext> </msup> </math></EquationSource> </InlineEquation>. Validated via phase portraits and time-series data, the results demonstrate satisfactory prediction accuracy even with a limited training dataset. Furthermore, comparing the proposed PINN model with traditional deep learning models for conflict-state prediction confirms that the proposed approach demonstrates significantly superior performance. This approach mitigates the sensitivity to the initial conditions in the fuzzy conflict model with external forces.</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

Status Prediction in a Fuzzy Conflict Model with Chaotic Behavior Using a Physics-Informed Neural Network

  • 期刊论文

  • 发表日期：

    2025年2月20日

查看全文
  • Gwiman Bak,
  • Youngchul Bae

摘要

Conflict is ubiquitous and unavoidable in human life and society. Moreover, predicting chaotic conflicts with sensitivity to the initial conditions generated by a fuzzy conflict model is challenging. This paper introduces a novel chaotic status prediction method for the fuzzy conflict model influenced by external forces, applying a physics-informed neural network (PINN), which integrates physical laws into machine learning, to address the challenges posed by limited data availability. The model is trained with only 20%, 40%, and 60% of the total data, and its performance is evaluated using the mean squared error, symmetric mean absolute percentage error, and \(R^{\textrm{2}}\) R 2 . Validated via phase portraits and time-series data, the results demonstrate satisfactory prediction accuracy even with a limited training dataset. Furthermore, comparing the proposed PINN model with traditional deep learning models for conflict-state prediction confirms that the proposed approach demonstrates significantly superior performance. This approach mitigates the sensitivity to the initial conditions in the fuzzy conflict model with external forces.