Unbalanced ensemble classification method based on data partition hybrid sampling

Abstract

The embodiment of the invention provides an unbalanced ensemble classification method based on data partition mixed sampling. The method comprises the following steps: dividing a sample space into four regions according to majority class proportions in minority class neighborhoods; generating a weight according to the ratio of the majority class ratio of each minority class neighborhood to the sumof the majority class ratios, the minority class safety regions, the boundary regions and the minority class noise regions, determining the synthesis number of each minority class neighborhood according to the weight, and performing oversampling on the minority classes of the boundary regions in a random linear interpolation mode; random under-sampling is carried out on the majority class of safety regions, a few class of noise region samples are removed, a few class of safety region samples are reserved, and a balance data set is generated; and constructing three ensemble learning models: anoriginal model biased to majority classes, a local domain reinforcement and weakening model and a hybrid model biased to peripheral boundaries, and adaptively selecting a corresponding model according to the unbalance degree of test point neighbors placed in an original data set.

Description

【Technical field】 [0001] The invention relates to a classification method for solving category imbalance in the field of machine learning, in particular to an unbalanced integrated classification method based on mixed sampling of data partitions. 【Background technique】 [0002] Classification has become a hot topic in machine learning, and it plays the role of data analysis and prediction in many application fields. For class imbalance, the data distribution of each class is uneven, and one or several classes (minority class) samples contain a small amount of data while other classes (majority class) samples contain a large amount of data. Given a training set with imbalanced categories, it is difficult to train an effective classification model, which is called an imbalanced classification problem. Many methods have been proposed to solve such problems, mainly divided into data-level methods, algorithm-level methods, and methods combining data processing and algorithms. T...

AI Technical Summary

Problems solved by technology

[0003] In view of this, the embodiment of the present invention proposes an unbalanced integrated classification method based on mixed sampling of data partitions, which can effectively solve the classification problem of unbalanced positive and negative samples, and generate different classification models by adjusting the data distribution to improve the unbalanced problem. Classification performance, improve the comprehensive performance of the classification model

Images