Support vector regression method for simultaneously optimizing mean square error and square multiple correlation coefficient

Abstract

The invention discloses a support vector regression method for simultaneously optimizing a mean square error and a square multiple correlation coefficient. The support vector regression method comprises the following specific steps: S1, segmenting and normalizing a data set; S2, initializing a support vector regression setting parameter group; S3, taking a mean square error and a square multiple correlation coefficient as multiple targets, and continuously updating a support vector regression setting parameter group; and S4, outputting optimized support vector regression setting parameters and corresponding mean square error and square multiple correlation coefficients. According to the method, the mean square error and the square multiple correlation coefficient are taken as multiple targets, the set parameters of support vector regression are optimized by using an iteration method, the mean square error is close to 0 as far as possible, and the square multiple correlation coefficient is close to 1 as far as possible, so that the problem of adjusting the set parameters of support vector regression is solved.

Description

technical field [0001] The invention relates to a support vector regression method for simultaneous optimization of mean square error and square complex correlation coefficient, belonging to the technical field of computers. Background technique [0002] Support vector regression is often used to build predictive models and is used in many application scenarios. However, there are many setting parameters. If you need to select a kernel function according to the feature number and capacity of the data set, after selecting the kernel function, you need to set the coefficient of the kernel function. Each kernel function has different coefficients, and these coefficients will bring about differences. The prediction results have a far-reaching impact. In addition, the impact of different coefficients on different targets is also different, especially the impact on the mean square error and squared complex correlation coefficient is rarely studied. [0003] Therefore, it is nove...

AI Technical Summary

Problems solved by technology

However, there are many setting parameters. If you need to select a kernel function according to the feature number and capacity of the data set, after selecting the kernel function, you need to set the coefficient of the kernel function. Each kernel function has different coefficients, and these coefficients will bring about differences. The predicted results of the far-reaching

In addition, the influence of different coefficients on different targets is also different, especially the influence on the mean square error and squared complex correlation coefficient is rarely studied

[0003] Therefore, it is novel to take the prediction output of support vector regression: mean square error and squared complex correlation coefficient as the target at the same time, and use iterative algorithm to integrate it into support vector regression, so that the setting parameters of support vector regression can be optimized. research is lacking

Images