Scholarlink

Feature selection is a critical step in building effective machine learning models, particularly in medical datasets where high-dimensional data can impede model performance and interpretability. This paper explores the application of large language models (LLMs), specifically Llama-based feature selection, and compares its performance against traditional techniques such as Recursive Feature Elimination (RFE) and KBest across various classification models. We evaluate models including AdaBoost, Decision Tree, Gradient Boosting, K-Nearest Neighbors (K-NN), Logistic Regression, Naive Bayes, Random Forest, and Support Vector Machine (SVM) on the Framingham heart disease dataset, analyzing their performance in terms of accuracy, recall, and precision. The results show that Llama-based feature selection methods generally outperform traditional techniques, with Logistic Regression and AdaBoost models exhibiting the most notable improvements. Our findings highlight the potential of LLM-based feature selection in enhancing predictive performance for medical datasets.

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

Leveraging Large Language Models for Zero Shot Feature Selection in Cardiometabolic Disease Prediction

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Kesanam Ashinee,
  • Gummuluri Venkata Ravi Ram,
  • M. Anand Kumar

摘要

Feature selection is a critical step in building effective machine learning models, particularly in medical datasets where high-dimensional data can impede model performance and interpretability. This paper explores the application of large language models (LLMs), specifically Llama-based feature selection, and compares its performance against traditional techniques such as Recursive Feature Elimination (RFE) and KBest across various classification models. We evaluate models including AdaBoost, Decision Tree, Gradient Boosting, K-Nearest Neighbors (K-NN), Logistic Regression, Naive Bayes, Random Forest, and Support Vector Machine (SVM) on the Framingham heart disease dataset, analyzing their performance in terms of accuracy, recall, and precision. The results show that Llama-based feature selection methods generally outperform traditional techniques, with Logistic Regression and AdaBoost models exhibiting the most notable improvements. Our findings highlight the potential of LLM-based feature selection in enhancing predictive performance for medical datasets.