VaR confidence interval prediction method

Abstract

The invention discloses a VaR confidence interval prediction method, and the method comprises the following steps: S1, carrying out the repeated sampling of initial sample data through a Bootstrap method, and generating a Bootstrap subsample; S2, carrying out gradual backward selection on the ARIMA model so as to calculate ARIMA model parameters, and extracting a random number sequence obeying an empirical distribution function according to the obtained ARIMA model parameters; S3, according to the Bootstrap subsamples and the random number sequence, obtaining a conditional variance estimation value of a T + 1 period sample prediction value XT + 1, and calculating a VaR prediction value at a T + 1 moment and a confidence interval of the VaR prediction value; S4, checking the VaR predicted value at the T + 1 moment and the confidence interval of the VaR predicted value; according to the method, the Bootstrap and the self-adaptive ARIMA model are combined to research the VaR predicted value, and the VaR confidence interval is verified, so that the fluctuation characteristics of the thick tail skewness of the financial market can be well described, the confidence interval of the statistical magnitude can be constructed based on empirical distribution, and higher prediction precision is obtained.

Description

technical field [0001] The invention relates to the technical field of risk measurement, in particular to a VaR confidence interval prediction method combining Bootstrap resampling and an adaptive ARIMA model. Background technique [0002] At present, the widely used risk measurement method in the financial market is VaR. The so-called VaR refers to the maximum value loss of financial assets or investment portfolios that may occur in a certain period of time in the future under a certain confidence level. This indicator can accurately measure the maximum possible loss and evaluate different risk factors. attention of financial institutions. However, the basic assumption of the traditional VaR calculation method is that the data sequence obeys a single distribution, such as a normal distribution. But in practice, many financial markets (such as stock and securities markets) are the same, and often have the characteristics of thick tail, heteroscedasticity, long memory and s...

AI Technical Summary

Problems solved by technology

However, there are still some deficiencies, most of which focus on the point prediction of market risk measurement, that is, use sample observations to study VaR, but do not discuss the VaR confidence interval under the existence of sampling error, which is the second type of problem to be solved

Images