Inferência para modelos de fração de cura zeros inflacionados aplicados a dados de risco de crédito

Abstract

Given the great demand for granting credit and service goods, there is a need to be able to control the risk involved in the process. This measure aims to manage possible unwanted events, for example, default, in order to enable the generation of profit or control losses so that they are not greater than what the financial institution could bear. In times of crisis, it becomes increasingly necessary to use tools that can help decision making more reliably. Thus, several statistical techniques are used to build models that can express risk scenarios, including survival analysis, which aims, for example, to predict situations such as the time until defaulting individuals return to their initial defaulting status ( credit recovery). With the application of these techniques, financial institutions can base themselves on the results in order to provide an ideal credit value, so that it does not generate losses for the same, as well as estimates for the resumption of credit operations. In this context, this work aims to study the survival model with a zero-inflated cure fraction. In this approach, it is possible to incorporate three classes of individuals: individuals with time equal to zero, non-susceptible and susceptible to the event of interest. The proposed methodology is applied to a database of a financial company.