Study Of Rare Events Based On Mixed Copula And Logistic Regression

Rare events rarely occur, but once they occur, a great impact will be produced, so the study of rare events has important practical significance. Compared with single copula function, mixed copula function can freely choose different types of copula function to create correlation structures, and it can accurately describe the complex relationship of rare events. The mixed copula function constructed by gumbel copula, clayton copula and frank copula, can not only cover the situation on the single upper tail or lower tail, but also show the situation on the upper tail and lower tail coexist. This paper uses the rare events logistic regression based on mixed copula function for the first time, and at the same time the model is added in the influence factors and associated events of rare events. As a result,this model can use the conditional probability to estimate the probability of rare events.Furthermore, it takes full advantage of copula functions in the tail dependence and logistic regression in the treatment of categorical variables, optimizing the probability estimation and the fitting precision of rare events.