Intelligent classification method based on support vector data description algorithm with privileged information

Abstract

The invention relates to an intelligent classification method based on a support vector data description algorithm with privileged information. The method includes the following steps: step 1, introducing the privileged information under an LUPI framework; step 2, associating the privileged information with relaxation factors, which means mapping novel training samples with the privileged information to a high-dimensional feature space by using nonlinear mapping; step 3, further obtaining an optimization equation of an SVDD+ with the privileged information according to an SVDD principle; step4, solving the optimization equation of the SVDD+ by using a Lagrange multiplier method to obtain a decision hypersphere; and step 5, performing prediction classification on a to-be-predicted sample by using the decision hypersphere. On the basis of the support vector data description algorithm, the privileged information of the sample is used to optimize the algorithm, which reflects internal relations of the mode. The intelligent classification method is an intelligent classifier design method with better overall performance, wider application field and more practical recognition result.

Description

technical field [0001] The invention relates to the field of intelligent classifiers in pattern recognition, in particular to an intelligent classification method based on the introduction of hidden information support vector data description algorithm. Background technique [0002] At present, classic KNN, SVM, ELM, decision tree, neural network, and support vector data description (SVDD) and other classification algorithms and their improvement schemes have achieved good classification results in some specific application fields. [0003] Among them, the support vector data description algorithm (SVDD) as the basis of the algorithm of the present invention is a supervised single classification algorithm proposed by Tax et al. in 1999, which has been widely used in abnormal behavior detection, abnormal identification and other fields. At present, the most commonly used method to improve the performance of SVDD classification is to optimize the boundary of the SVDD hypersphe...

AI Technical Summary

Problems solved by technology

[0008] In order to solve the problems that the traditional classifiers in the existing technology cannot provide satisfactory processing results for the large number of labels, difficulties in target recognition and outlier detection, poor data separability, and some hidden information in the data set cannot be effectively used , still can cause the technical problem that classification method accuracy is not high, the present invention provides following technical scheme:

Images