基于机器学习与生存模型建立脂肪肝Joint联合预测模型

目的 利用机器学习算法与生存模型建立脂肪肝Joint联合预测模型，为有关单位进行脂肪肝健康管理提供理论依据。方法 选取2006—2016年某人群体检数据为研究对象，据模拟实验结果选择机器学习方法建立纵向亚模型，利用时依Cox模型建立生存亚模型，再联合建模。结果 XGBoost算法F-measure值最大，均方误差最小，建立纵向亚模型。XGBoost-Joint联合模型稳定性和拟合效果优于其他组合的Joint模型。结论 Joint联合模型将纵向亚模型与生存亚模型相联系，关联变量在模型中对脂肪肝患病风险的影响明显上升，说明纵向过程对生存结果的影响很大。

Joint combined forecast model of fatty liver was established based on machine learning and survival model

To establish Joint combined forecast model of fatty liver with machine learning algorithm and survivalmodel，and to provide theoretical basis for health management of fatty liver in related units. Methods The physicalexamination data of a certain population from 2006 to 2016 were selected as the research object. According to the results ofsimulation experiment，machine learning method was selected to establish the longitudinal submodel，and the survival submodelwas established by using time － dependent Cox regression model，and then the combined model was conducted. Ｒesults TheXGBoost algorithm had the maximum F － measure value and the minimum mean square error，and the longitudinal submodelwas established. The stability and fitting effect of XGBoost － Joint combined forecast model were better than those of other Jointmodels. Conclusion Joint combined forecast model connects the longitudinal submodel and the survival submodel，theassociated variables of the Joint combined forecast model significantly increase the influence on the risk of fatty liver disease，indicating that the longitudinal process has a great influence on the survival outcome.