PhysioNet Challenge 2012

该数据集包含了来自重症监护室（ICU）的12,000个多变量临床时间序列记录，特别使用了训练集A，其中包含了4,000次ICU住院数据。该数据集的缺失率高达80.53%，且数据非常稀疏，这给诸如预测患者院内死亡等下游任务带来了复杂性。数据集的规模为192,000个样本和37个变量，其任务是进行患者特异性的院内死亡预测。This dataset contains 12,000 multivariate clinical time-series records collected from intensive care units (ICUs). Specifically, Training Set A is utilized, which includes 4,000 ICU admissions. The dataset has a missing rate as high as 80.53% and is extremely sparse, introducing complexity to downstream tasks such as predicting patients' in-hospital mortality. It consists of 192,000 samples and 37 variables, with the core task being patient-specific in-hospital mortality prediction.

搜集汇总

背景与挑战

背景概述

PhysioNet Challenge 2012数据集包含12,000名成人ICU患者的临床记录，用于开发院内死亡率预测模型。数据涵盖42个变量，包括人口统计学信息、生命体征和实验室检查的时间序列数据，时间跨度为入院后48小时。数据集已划分为训练集和测试集，支持二元分类和风险估计两种预测任务，是医疗预测建模的基准数据集。