Scholarlink

<p>Rapid advancements in medical technology have enabled the efficient acquisition of large volumes of diverse medical data. However, challenges exist in effectively extracting disease-relevant knowledge or predictive features from such data to support clinical decision-making. Therefore, this study developed a framework that integrates multiple machine learning models and optimization algorithms to enable the early prediction of cardiovascular disease, Parkinson’s disease (PD), and nonalcoholic fatty liver disease (NAFLD). Datasets related to these diseases were used to train three machine learning models, namely categorical boosting, light gradient boosting machine, and tabular neural network. To enhance the prediction accuracy and decrease the error rates of these models, hyperparameter optimization was performed using various optimization algorithms. The experimental results indicate that after hyperparameter optimization, the highest prediction accuracies for the detection of cardiovascular disease, PD, and NAFLD reached 85.28%, 97.44%, and 93.08%, respectively, with relatively small differences observed in the performance achieved with different model–algorithm combinations. In addition, the model performance data collected in this study were input to a large language model, which was integrated with a retrieval-augmented generation framework to develop a medical recommendation system. When model performance data were uploaded to this system, it could automatically identify the disease, recommend the optimal model–algorithm combination for prediction, and provide relevant medical advice. This system offers a reliable method for early disease prediction and has high potential for adoption in health-care applications driven by artificial intelligence.</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

An intelligent disease prediction framework integrating machine learning, optimization, and LLM with RAG

  • 期刊论文

  • 发表日期：

    2026年7月29日

查看全文
  • Ping-Huan Kuo,
  • Ping-Chun Tsai,
  • Bang-Yu Chen,
  • Ting-Chun Yao

摘要

Rapid advancements in medical technology have enabled the efficient acquisition of large volumes of diverse medical data. However, challenges exist in effectively extracting disease-relevant knowledge or predictive features from such data to support clinical decision-making. Therefore, this study developed a framework that integrates multiple machine learning models and optimization algorithms to enable the early prediction of cardiovascular disease, Parkinson’s disease (PD), and nonalcoholic fatty liver disease (NAFLD). Datasets related to these diseases were used to train three machine learning models, namely categorical boosting, light gradient boosting machine, and tabular neural network. To enhance the prediction accuracy and decrease the error rates of these models, hyperparameter optimization was performed using various optimization algorithms. The experimental results indicate that after hyperparameter optimization, the highest prediction accuracies for the detection of cardiovascular disease, PD, and NAFLD reached 85.28%, 97.44%, and 93.08%, respectively, with relatively small differences observed in the performance achieved with different model–algorithm combinations. In addition, the model performance data collected in this study were input to a large language model, which was integrated with a retrieval-augmented generation framework to develop a medical recommendation system. When model performance data were uploaded to this system, it could automatically identify the disease, recommend the optimal model–algorithm combination for prediction, and provide relevant medical advice. This system offers a reliable method for early disease prediction and has high potential for adoption in health-care applications driven by artificial intelligence.