Application method of fuzzy support vector machine in telephone traffic prediction

Abstract

The invention relates to an application method of a fuzzy support vector machine in telephone traffic prediction, and aims to obtain a new kernel function to be applied in the telephone traffic prediction by adding priori knowledge described by a linguistic variable or fuzzy set to a design of the kernel function. The application method comprises the following steps of: firstly obtaining a membership function in a first component of a fuzzy rule of a TSK (Takagi-Sugeno-Kang) model by a fuzzy C-means clustering algorithm; secondly identifying a second component parameter of the TSK model by an epsilon insensitive loss function; and finally obtaining a fuzzy membership kernel function through a current kernel function and the membership function or probability distribution function by a kernel trick. According to the application method of the fuzzy support vector machine in the telephone traffic prediction, provided by the invention, a great number of comparison experiments are developed by taking a mean-square error as a performance index for balancing a regression effect, thus the FMK (Fuzzy Membership Kernel Function) has better robustness.

Description

technical field [0001] The invention relates to an application method of a fuzzy support vector machine in traffic prediction. Background technique [0002] Based on the fuzzy Takagi-Sugeno-Kang (TSK) model and the ε-insensitive loss function, a new kind of kernel function is constructed by using the fuzzy membership function and the existing kernel function through the kernel technique, that is, the fuzzy membership kernel function (FMK ). Firstly, the fuzzy membership function is used to represent prior knowledge, and then the characteristics of prior knowledge are combined into the construction process of kernel function according to fuzzy reasoning, so as to improve the performance of kernel function. Secondly, FMK is applied to the support vector regression machine, and a large number of comparative experiments show that FMK has good performance. On the one hand, comparing FMK with polynomial kernel, Gaussian kernel and spline kernel function, the experimental results...

AI Technical Summary

Problems solved by technology

Existing methods are based on certain assumptions about the training set, using deterministic prior knowledge, but cannot be integrated into the design process of the kernel function

In real life, people often use inaccurate, incomplete or incompletely reliable information for reasoning, that is, uncertainty reasoning, which limits the application of kernel function methods.

Images