Nonlinear Granger causality detection method based on kernel recursion maximum cross-correlation entropy algorithm

Abstract

The invention discloses a nonlinear Granger causality detection method based on a kernel recursion maximum cross-correlation entropy algorithm. The method is based on a traditional Granger causality detection method. A kernel recursion maximum cross-correlation entropy algorithm is adopted to establish a nonlinear regression model. A nonlinear regression error is utilized to calculate a causal index. By acquiring a causal relationship between two time sequences through a causal criterion, and significance testing is carried out on the causal index and a resolving power index. According to themethod, the nonlinear causal relationship can be correctly detected, and the resolution capability is higher. Therefore, the nonlinear Granger causality detection method based on the kernel recursionmaximum cross-correlation entropy algorithm is easier to popularize and use in practical application.

Description

technical field [0001] The invention belongs to the field of time series analysis, and in particular relates to a nonlinear Granger causality detection method based on a kernel recursive maximum cross-correlation entropy algorithm. Background technique [0002] Causality (unlike correlation, which describes the degree of interaction between variables and has no directionality) is the relationship between one event (ie, "cause") and a second event (ie, "effect"), where the latter Events are considered to be the result of previous events and have directionality, so causality detection is important to describe the relationship between variables. Granger causality (Granger causality, GC) is the most commonly used causality test method. It is simple to calculate and easy to understand. It has been successfully applied in many fields, such as economics, climate research, genetics and neuroscience. However, it is based on a linear regression model and cannot be used to test the ca...

AI Technical Summary

Problems solved by technology

However, it is based on a linear regression model and cannot be used to test the causality of time series with nonlinear causality; second, the linear regression model it uses is based on the minimum mean square error criterion, when the signal is heavily polluted by non-Gaussian noise , linear regression has poor accuracy and may detect false causal relationships

Images