Method for selecting kernel function of support vector machine based on sample prior information and application

Abstract

The invention relates to a method for selecting a kernel function of a support vector machine based on sample date prior information, and is particularly applicable to real-time online support vector machine model prediction control places. The method comprises the following steps: inputting sample data, wherein Rn is n-dimensional data space, and X is converted to enable the norm of the data to be less than 1; performing hypersphere mathematic description on the given sample data and determining the gravity center O and the radius R of the hypersphere; establishing a sample distribution energy entropy function, and calculating the energy entropy of each sample; constructing a sample distribution discrimination function and calculating the discrimination result of the function; selecting the kernel function type according to the similarity between the discrimination result and the kernel function properties (such as Riemann metric and distance measurement); reasonably determining a sample training set and a testing set, and optimizing the SVM model and parameters; and outputting a prediction result. With the method, the SVM studying ability and generalization ability are improved, and the method has the characteristics of high operation speed and the like, and is particularly suitable for real-time online SVM model prediction control places.

Description

Technical field [0001] The invention relates to a support vector machine kernel function selection method and application based on prior information of sample data, and is particularly suitable for real-time online support vector machine model predictive control sites. Background technique [0002] Support Vector Machine (SVM) is a new machine learning method proposed by Vapnik based on statistical learning theory in the 1990s. Compared with traditional statistics, support vector machines have a complete theoretical foundation and a strict theoretical system, which can solve learning problems with limited samples and have strong generalization capabilities. Because this method has many excellent characteristics, it has been successfully applied in many fields such as pattern recognition, regression estimation, data mining, and bioinformatics. SVM is based on the principle of structural risk minimization. One of its core ideas is to introduce kernel function technology, which cle...

AI Technical Summary

Problems solved by technology

However, different kernel functions exhibit different characteristics, and choosing different kernel functions will lead to different generalization performance of SVM

At present, how to select (or construct) a suitable kernel function for a specific problem lacks corresponding theoretical guidance, and there is a lot of randomness and limitations, which is also a major problem encountered in the field of SVM applications.