基于多模态数据的救援人员PTSD风险因素识别与预测研究

Research on Risk Factor Identification and Prediction of Post-Traumatic Stress Disorder (PTSD) in Rescue Workers Based on Multimodal Data

摘要: 消防救援人员长期执行高强度、高风险任务，频繁接触生死与灾后惨烈场景，因而心理负担极重，罹患创伤后应激障碍(PTSD)的风险显著高于普通人群，针对这一高危职业群体开展PTSD影响因素研究，具有重要的理论价值和实践意义。本文基于多模态数据，非结构化数据采用快速傅里叶变换选择特征，使用ROSE方法对训练集进行平衡处理，利用XGBoost内置的特征重要性评估机制对小样本高维数据建立个体化救援人员PTSD预测模型，并在独立测试集上验证得到模型的稳健性，为构建精准化心理健康管理体系提供了可推广的范式。

Abstract: Fire and rescue personnel frequently perform high-intensity, high-risk tasks and are often exposed to life-and-death situations and horrific scenes after disasters. Consequently, they experience significant psychological burden and have a significantly higher risk of developing post-traumatic stress disorder (PTSD) than the general population. Conducting research on the influencing factors of PTSD in this high-risk occupational group has important theoretical and practical significance. This paper, based on multimodal data, uses Fast Fourier Transform to select features from unstructured data, employs the ROSE method to balance the training set, and utilizes XGBoost’s built-in feature importance evaluation mechanism to build an individualized post-traumatic stress disorder (PTSD) in rescue workers prediction model for rescue personnel using small-sample, high-dimensional data. The model’s robustness is then validated on an independent test set, providing a generalizable paradigm for constructing a precise mental health management system.