Scholarlink

<p>To address the limited adaptability of existing causal reasoning methods, this study proposes a causal reasoning model based on Continual Learning and an improved Bayesian network, combined with case-based model analysis. The study optimizes the input of the K2 structure learning algorithm through the Most Weight Supported Tree algorithm and Ant Colony Optimization, conducts causal analysis according to the learned structure, and then performs accident severity reasoning using parameter learning based on Maximum Likelihood Estimation, with adaptive adjustment through Continual Learning. In addition, the study applies the model to four different scenarios, designing practical experiments to analyze the model’s performance. In the practical experiments, the proposed model achieves an average absolute error of 3.36% in scenario 1 and an average relative error of 3.02% in scenario 3, both outperforming comparative models and demonstrating superior predictive performance. The experimental results indicate that the proposed model has high accuracy and stability, efficiently conducts causal analysis and prediction, and promotes the development of causal reasoning.</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

Instantiated causal reasoning based on machine learning

  • 期刊论文

  • 发表日期：

    2026年6月17日

查看全文
  • Zengguang Wang

摘要

To address the limited adaptability of existing causal reasoning methods, this study proposes a causal reasoning model based on Continual Learning and an improved Bayesian network, combined with case-based model analysis. The study optimizes the input of the K2 structure learning algorithm through the Most Weight Supported Tree algorithm and Ant Colony Optimization, conducts causal analysis according to the learned structure, and then performs accident severity reasoning using parameter learning based on Maximum Likelihood Estimation, with adaptive adjustment through Continual Learning. In addition, the study applies the model to four different scenarios, designing practical experiments to analyze the model’s performance. In the practical experiments, the proposed model achieves an average absolute error of 3.36% in scenario 1 and an average relative error of 3.02% in scenario 3, both outperforming comparative models and demonstrating superior predictive performance. The experimental results indicate that the proposed model has high accuracy and stability, efficiently conducts causal analysis and prediction, and promotes the development of causal reasoning.