Survival analysis is widely used in medical,economic,industrial,social sciences and other fields.Survival analysis is used to study the statistical relationship between survival phenomenon,survival time and its influencing factors.Many traditional statistical methods have been widely used in the field of survival analysis.In recent years,with the rapid development of machine learning algorithms,many algorithms have been introduced into the survival analysis,and their superior modeling ability and prediction performance have been verified in practice.The objective of this thesis is to propose an efficient regression algorithm of survival analysis-Survival Boost.This algorithm is based on Random Survival Forests(RSF)and XGBoost.By combining the Elastic-Net penalty type Cox proportional hazards regression model with XGBoost optimal algorithm,our algorithm is more suitable for survival analysis.The performance of Survival Boost is compared with the Cox model,XGBoost,Cox Boost,RSF and Gradient Boosting Desicion Tree model on 4 simulated datasets and 4 real survival datasets.The results illustrated the superiority of the algorithm.This thesis uses Shapley Additive Explanation values(SHAP values)to explain the characteristics of the model,which further explains the effectiveness of Survival Boost to guide the diagnosis and practice. |