METHOD AND APPARATUS FOR MULTI-CLASS CLASSIFICATION USING SUPPORT VECTOR DOMAIN DESCRIPTION, AND COMPUTER-READABLE STORAGE MEDIUM USED THERETO

摘要

A method and a device for performing multi-class classification using SVDD(Support Vector Domain Description), and a computer-readable storage medium for the same are provided to show a correct classification prediction for the multi-class classification, apply to unbalanced data, and offer class probability by inferring a pseudo-density function of data divided into each class through the SVDD, which is found a posterior probability function based on the pseudo-density function, and classifying the data based on a Bayesian optimal decision theory. A data dividing module divides the data into datasets(S100). A Gaussian kernel supporting function extracting module extracts Gaussian kernel supporting functions by applying SVDD to each data set(S110). A pseudo-density function inferring module infers a pseudo-density function for each class(S120). A data classifying module classifies the data by applying a Bayesian optimal decision theory with using the inferred pseudo-density function as a class-conditional density function(S130). A predicting module provides posterior probability for the data inputted in future and classifies the data into the classes by storing a result model found through the data classifying module(S140).