Credit rating default probability measurement and risk early warning method

Abstract

The invention belongs to the technical field of financial information data management, and discloses a credit rating default probability measure and risk early warning method. The method comprises thefollowing steps: mining valid data and measuring and calculating default probability; constructing a default loss rate prediction model, and checking the consistency of rating results; and constructing a binary response risk early warning model, and realizing an early warning strategy by means of a data mining technology. Through research and application of a credit rating mathematical model, a set of credit rating system which can be verified by using data and is high in feasibility is finally formed by using a data mining technology, and effective practical experience is obtained. Meanwhile, the research result can cooperate with the financial industry and can also cooperate with credit enterprises to provide credit products and technical services for the enterprises and provide data reference for transaction decisions of market subjects, so that contribution is made to credit demands and credit industry development.

Description

technical field [0001] The invention belongs to the technical field of financial information data management, and in particular relates to a credit rating default probability measurement and risk early warning method. Background technique [0002] Currently, the closest prior art: [0003] The core issue of credit rating is the study of the probability of default, which is divided into two aspects: the measurement of the probability of default and the evaluation of the probability of default. The former is to solve the problem of how to measure the probability of default; the purpose of the latter research is to analyze the relevant factors and their importance of decision, influence default and default probability. [0004] The research on the measurement of default probability can be classified into the following categories: credit rating default probability based on credit rating historical data, "base" default rate based on option pricing theory, measurement of default ...

AI Technical Summary

Problems solved by technology

[0009] 2. Lack of consistency test for rating results Since Louis Tappan, the world's first commercial credit rating agency, was established in New York in 1841, raters, statisticians, mathematicians and software programmers have created countless credit rating models and developed many credit rating models. There are many fast rating software, and my country's rating agencies have also introduced a lot of advanced rating software, but in the actual credit rating work, people pay more attention to the rating results, but lack of consistency check and correction of the rating results

[0018] (1) The network structure of the neural network model is difficult to determine and overfit, which largely affects its ability to predict new sample defaults and the application of the model

[0019] (2) There is a difference between the default rate calculated based on historical credit data and the actual default probability in the same period, and the fitting between the real data and the original default probability measurement model is not ideal

[0020] (3) In the actual credit rating work, people pay more attention to the rating results, but lack the consistency check and correction of the rating results

[0021] (4) Serious lack of data in the later period, resulting in difficulties in tracking and rating work, low timeliness of rating information, and lagging risk early warning

[0022] (5) The original mathematical model rarely considers the randomness of variables in the financial market, and the rating system, method or actual operation of the rating agency is not stable enough;

[0023] (6) Part of the actual credit rating work is qualitative index factors. If the impact of qualitative factors on ratings is ignored, the accuracy of default probability measurement will be reduced;

[0024] (7) The principal component analysis algorithm will lose part of the original data information in the process of dimensionality reduction. After the data processing of the principal component analysis algorithm, if the contribution rate accounts for more than 85%, then the system will consider the subsequent data results to have statistical value. However, the missing data information directly affects the accuracy and reliability of ratings;

Images