讲座题目:生存分析与违约概率预测(On the application of survival analysis to forecasting the probability of default)
主讲人:张晓奇讲师
讲座时间:2017年11月9日下午2:00-4:00
讲座地点:BV伟德国际体育楼422会议室
讲座摘要:Identifying the factors that could affect the probability of default (PD) is the central task in both of the academic studies and industrial practices of financial risk management. There are multiple empirical methods to analyze PD, including the logistic regression, machine learning, survival analysis and the like. In this talk, we will focus on the application of survival analysis methods to explaining and forecasting PD in multiple settings, like the PD of commercial bonds and the counter-party PD. I will give an overview of the basic concepts in survival analysis and present their connection to the studies of PD. I will also briefly review the pros and cons of the survival analysis methods comparing to the logistic regression and machine learning methods. The models we will cover in this talk include the Cox proportional hazard model, the Cox model with temporal covariates and the joint model of longitudinal measurements and failure time. Through a comprehensive comparison among the three classes of models, we claim that the joint model of longitudinal measurements and failure time might be the best fit to the forecast purpose. Finally, I will discuss the possibility to apply the joint model method to some real PD data.