ClarusC64/clinical-hemostasis-instability-v0.1

该数据集用于评估模型是否能检测止血调节中的不稳定性。每一行代表一个简化的凝血监测场景，观察三个时间点。任务是确定凝血调节是否保持稳定或趋向于止血不稳定性。止血稳定性取决于血凝块形成和分解之间的平衡。涉及信号包括血小板计数轨迹、D-二聚体代理轨迹、凝血激活代理、纤维蛋白溶解代理轨迹、血管损伤代理和干预延迟。当凝血激活增加而纤维蛋白溶解平衡和血小板调节恶化时，不稳定性出现。预测目标：label = 1表示止血不稳定性，label = 0表示稳定的凝血平衡。每行包括血小板计数轨迹、D-二聚体代理轨迹、凝血激活代理轨迹、纤维蛋白溶解代理轨迹、血管损伤代理和干预延迟。还包括干扰变量：实验室噪声和图表噪声。评估方法要求预测遵循特定格式，并输出准确率、精确率、召回率、F1分数、混淆矩阵和数据集完整性诊断。

This dataset evaluates whether models can detect instability in hemostatic regulation. Each row represents a simplified coagulation monitoring scenario observed across three time points. The task is to determine whether coagulation regulation remains stable or is moving toward hemostatic instability. Hemostatic stability depends on balance between clot formation and clot breakdown. Signals that interact include: platelet count trajectory, D-dimer proxy trajectory, coagulation activation proxy, fibrinolysis proxy trajectory, vascular injury proxy, and intervention delay. Instability emerges when coagulation activation increases while fibrinolytic balance and platelet regulation deteriorate. Prediction target: label = 1 → hemostatic instability, label = 0 → stable coagulation balance. Each row includes: platelet count trajectory, D-dimer proxy trajectory, coagulation activation proxy trajectory, fibrinolysis proxy trajectory, vascular injury proxy, and intervention delay. Decoy variables: lab_noise, chart_noise. Evaluation requires predictions to follow a specific format and outputs accuracy, precision, recall, f1, confusion matrix, and dataset integrity diagnostics.