ClarusC64/clinical-sepsis-trajectory-instability-v0.1

该数据集用于测试模型是否能够从短期临床代理序列中分类败血症轨迹的不稳定性。每个数据行描述了一个类似患者的场景，跨越三个时间点。任务是根据这些场景预测患者是趋向不稳定还是保持稳定。数据集的核心思想是败血症的不稳定性不仅仅依赖于单一变量，而是多个变量的交互作用，包括温度轨迹、血压轨迹、乳酸轨迹、炎症负担、呼吸压力、液体反应、干预延迟和肾脏压力等。预测目标为label=1表示败血症轨迹不稳定，label=0表示稳定或恢复轨迹。数据集还包括干扰变量，这些变量看似有意义但不单独决定目标。数据集评估指标包括准确率、精确率、召回率、F1分数和混淆矩阵等。

This dataset tests whether a model can classify sepsis trajectory instability from short clinical proxy sequences. Each row describes a patient-like scenario across three time points. The task is to predict whether the scenario is moving toward instability or remaining stable. The core idea is that sepsis instability does not depend on one variable alone but on interactions across multiple variables including temperature trajectory, blood pressure trajectory, lactate trajectory, inflammatory burden, respiratory strain, fluid response, intervention delay, and renal stress. The prediction target is label=1 for sepsis trajectory instability and label=0 for stable or recovering trajectory. The dataset also includes decoy variables that appear meaningful but do not define the target alone. Evaluation metrics include accuracy, precision, recall, F1 score, and confusion matrix.