预先训练有素的大型语言模型的表现优于统计和机器学习预测在急救部门的访问
Yi-Chang Yen1, Chin-Chieh Wu2, Shu-Hui Chen3
1Department of Medical Education, Chang Gung Memorial Hospital, Kaohsiung, Taiwan.
The American journal of emergency medicine
|September 13, 2025
概括
大型语言模型 (LLM) 在预测每日急诊室 (ED) 访问方面表现优异,优于传统方法. 这些先进的模型为预测患者流动提供了更高的准确性,即使是在COVID-19大流行期间.
科学领域:
- 医疗分析 医疗分析
- 时间序列预测时间序列预测
- 医学中的人工智能.
背景情况:
- 紧急诊所 (ED) 拥挤是一个持续的挑战,往往由不可预测的患者数量加剧.
- 准确预测每日ED访问对于优化资源分配和减轻过度拥挤至关重要.
- 随着COVID-19的流行,出现了显著的波动,影响了传统的预测模型.
研究的目的:
- 为了比较统计,机器学习和大语言模型 (LLM) 在预测每日ED访问中的有效性.
- 评估台湾在疫情前,疫情前和疫情后期间的模型性能.
- 为各种医院设置确定最准确的预测方法.
主要方法:
- 利用了6家台湾医院的每日ED访问记录 (2007-2022年),包括日历数据和COVID-19指标.
- 开发和比较统计模型 (SARIMAX,Prophet),机器学习模型 (LightGBM,LSTM,DLinear,TiDE) 和预先训练的基于变压器的LLM.
- 采用了七天的预测时间,并使用平均绝对百分比误差 (MAPE) 评估业绩.
主要成果:
- 预先培训的LLM获得了最好的整体表现 (MAPE 7.59%),紧随其后的是LightGBM (MAPE 8.08%).
- 在COVID-19之前的时期,Prophet (MAPE 6.80%) 和TiDE (MAPE 5.89%) 的表现非常出色.
- 尽管在COVID-19期间的表现下降,但LLM和LightGBM表现出了弹性.
结论:
- 预先训练有素的LLM为ED访问预测提供了卓越的整体准确性,特别是在流行病等波动时期.
- 轻GBM在不同的时间框架中提供了强大的性能.
- 先进的时间序列模型,包括LLM,具有提高ED运营效率的巨大潜力.
更多相关视频
05:47Evidence-based Knowledge Synthesis and Hypothesis Validation: Navigating Biomedical Knowledge Bases via Explainable AI and Agentic Systems
Published on: June 13, 2025
1.3K
07:31Implementation of a Real-Time Psychosis Risk Detection and Alerting System Based on Electronic Health Records using CogStack
Published on: May 15, 2020
7.5K
相关概念视频
Steps in Outbreak Investigation
492
In the ever-evolving field of public health, statistical analysis serves as a cornerstone for understanding and managing disease outbreaks. By leveraging various statistical tools, health professionals can predict potential outbreaks, analyze ongoing situations, and devise effective responses to mitigate impact. For that to happen, there are a few possible stages of the analysis:
492
Statistical Methods for Analyzing Epidemiological Data
900
Epidemiological data primarily involves information on specific populations' occurrence, distribution, and determinants of health and diseases. This data is crucial for understanding disease patterns and impacts, aiding public health decision-making and disease prevention strategies. The analysis of epidemiological data employs various statistical methods to interpret health-related data effectively. Here are some commonly used methods:
900
Types of Reports III: Telephone and Verbal Reports
987
Telephone and Verbal Reports in healthcare settings are two communication methods for conveying therapeutic instructions from healthcare providers to nurses or other healthcare staff.
Here's an overview of each type:
Telephone Orders
Here's an overview of each type:
Telephone Orders
987
Survival Tree
389
Survival trees are a non-parametric method used in survival analysis to model the relationship between a set of covariates and the time until an event of interest occurs, often referred to as the "time-to-event" or "survival time." This method is particularly useful when dealing with censored data, where the event has not occurred for some individuals by the end of the study period, or when the exact time of the event is unknown.
Building a Survival Tree
Constructing a...
Building a Survival Tree
Constructing a...
389
