A method and system for classifying unbalanced data sets

Abstract

The method and system for classifying unbalanced data sets of the present invention calculate and obtain the class centers c of positive and negative training sets 1 and c 2 , determine the distance T of the two types of class centers, the positive class hyperplane, the negative class hyperplane, the first distance, the second distance, the third distance and the fourth distance, and then determine the fuzzy membership function. According to the fuzzy membership function and fuzzy A dual support vector machine determines the classification model. The optimized classification model is obtained by grid search algorithm and cross-validation method. Input the unbalanced data to be classified into the optimized classification model, and obtain the classification result of the unbalanced data to be classified. The method or system of the present invention assigns different membership values ​​to the sample points according to the difference in the contribution of the sample points to the classification hyperplane and the difference in the unbalanced rate of the two types of samples by using a deterministic classification model based on the fuzzy membership function. Therefore, the accuracy of the classification results when using the method or system of the present invention is improved.

Description

technical field [0001] The invention relates to the technical field of unbalanced data processing, in particular to a method and system for classifying unbalanced data sets. Background technique [0002] Many industry data often have imbalanced data distribution. Taking the binary classification problem as an example, if the proportion of one kind of samples is much larger than the proportion of the other kind of samples, the data set is an unbalanced data set. Among them, the majority class samples are also called negative class samples, the minority class samples are called positive class samples, and the ratio of the number of negative class samples to positive class samples is called Imbalanced Rate (IR). Typical examples include: fault diagnosis data, credit fraud data, medical diagnosis data, etc. Since the classification prediction accuracy of the minority class is more important in practice when classifying and predicting an unbalanced dataset, the commonly used cl...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a method and system for classifying unbalanced data sets, so as to solve the problems of low efficiency and poor accuracy in classifying unbalanced data sets in the prior art

Images