Classification of class-imbalanced data

Abstract

The present invention relates to the data mining technology, and especially relates to a method for training a class-imbalanced data classifier, a class-imbalanced data classifier and a method for classifying the class-imbalanced data. According to one embodiment of the method for training the class-imbalanced data classifier, data classified by the class-imbalanced data classifier has a plurality of properties. The method comprises the following steps that the properties are divided into a plurality of property groups, each property group corresponds to one sub-classifier, and each sub-classifier is suitable for classifying the data based on the corresponding property group, so as to obtain an ultimate classification result by the classification results of the sub-classifiers according to pre-set rules; training data samples are divided into multiple test sets; and for each property group, the corresponding sub-classifiers are trained by using different test sets.

Description

field of invention [0001] The invention relates to data mining technology, in particular to a training method for an unbalanced data classifier, an unbalanced data classifier and a method for classifying unbalanced data. Background technique [0002] Classification is one of the most commonly used techniques in data mining and machine learning, in which a classifier is trained on a set of objects of known classes and then objects of unknown classes are applied to the classifier to determine the corresponding classes. In the unbalanced class data, the number of samples of a certain class is much larger than that of other classes, the former is called negative class data, and the latter is called positive class data. [0003] In practical applications (such as credit card transaction fraud detection, network intrusion detection, medical disease diagnosis, etc.), the classification problem of unbalanced data is often encountered. The common point of these problems is that minor...

AI Technical Summary

Problems solved by technology

For example, in the application of credit card transaction fraud detection, more attention is paid to fraudulent customers, but most of the monitored data sets are normal credit card transaction records, and the proportion of illegal transactions is very small

There are many existing techniques for dealing with classification problems, such as decision trees, Bayesian networks, support vector machines, etc., but these techniques are mostly designed for balanced data, without considering the huge difference in the distribution of positive and negative data nature, so the processing effect is not good

[0004] At present, the classification of unbalanced data mainly adopts two ideas: one is to change the distribution of training set samples to reduce the degree of imbalance, which mainly includes resampling methods that change the distribution of data sets. The disadvantage is that the classification effect depends on the resampling algorithm. For many applications, it is difficult to determine the optimal distribution of data sets; the second is to construct new algorithms or transform existing algorithms according to the characteristics of unbalanced data (such as cost-sensitive learning methods, feature selection methods, and single-class learning methods, etc.), The disadvantage of the cost-sensitive learning method is that it is difficult to give an accurate estimate of the cost of misclassification, so that the overall performance improvement cannot be guaranteed. Feature selection methods are more suitable for text classification problems, and the scope of application is relatively limited. The disadvantage of the single-class learning method is that only a small number of positive data are used, and the useful information contained in the negative data is completely ignored.

Images