Virtual sample generation method based on kernel density estimation and Copula function

Abstract

The invention discloses a virtual sample generation method based on kernel density estimation and a Copula function. The method includes: obtaining an original sample set and an original training set,constructing an initial classification model according to the original sample set and the training set, obtaining a probability density estimation function of the original sample set according to a kernel density estimation method and positive samples in the original sample set, obtaining Copula model parameters according to a maximum likelihood estimation method; and constructing a joint densityfunction of the positive samples according to the Copula model parameters, obtaining a virtual sample set through re-sampling by using the joint density function, and determining the generation number of the virtual sample set according to the difference between the data volume of the negative samples and the data volume of the positive samples in the original sample set. According to the technical scheme provided by the invention, different types of data distribution conditions of the original data set can be effectively improved, and the classification effect of various classifiers under the unbalanced sample condition can be improved, so that the generalization capability of the classifiers is improved.

Description

Technical field [0001] The present invention relates to the technical field of machine learning, in particular to a virtual sample generation method based on kernel density estimation and Copula function. Background technique [0002] Pattern classification is one of the most basic intelligences that humans are born with. Since the day when people try to show intelligence on computers, pattern classification has naturally become the main research issue. In recent years, with the continuous development of the computer field, more and more excellent classification algorithms have emerged, such as Decision Tree (DT), Support Vector Machine (SVM), and K proximity algorithm (k -Nearest Neighbor, knn) etc. These classification algorithms have greatly improved the level of computer pattern classification, and can approach or even reach the recognition level of humans in many fields. However, classification algorithms often have high requirements for training samples, usually requiring...

AI Technical Summary

Problems solved by technology

When traditional machine learning methods are used to solve these unbalanced classification problems, the performance of the classifier often drops sharply, and the resulting classifier has a large bias

Images