Method and device for constructing unbalanced sample classification model

Abstract

The present disclosure provides a method for constructing an imbalanced sample classification model. The method includes receiving raw sample population data, constructing a classification model, and outputting the classification model. Wherein, constructing the classification model includes calculating the k closest first sample individuals in the minority class for each first sample individual belonging to the minority class in the original sample population, where k is greater than or equal to is a positive integer of 2, the classification of all sample individuals in the original sample population is known; then for each first sample individual, at least one is determined according to the k nearest first sample individuals second sample individual; then mixing the original sample population with all the second sample individuals to form a mixed sample population; finally constructing a classification model based on the mixed sample population. The present disclosure also provides a device, a system and a readable storage medium for constructing an imbalanced sample classification model.

Description

technical field [0001] The present disclosure relates to the field of computer technology, and more specifically, to a method and device for constructing an imbalanced sample classification model. Background technique [0002] In data mining classification or prediction tasks, sometimes the obtained data distribution is not balanced. For the unbalanced samples with a large disparity in the proportion of each category, the characteristics of the minority category with less data are easy to be ignored, which makes it easy to classify the data that should belong to the minority category into the majority category when making data predictions. This is very unfavorable for the analysis of unbalanced samples with a small minority class but a very large influence. For example, when network users are divided into normal users and malicious users (such as network hackers), the number of malicious users is far smaller than that of normal users, but the destructive power of these mali...

AI Technical Summary

Problems solved by technology

Oversampling will cause overfitting problems in the model and affect the performance of the classification model

Undersampling may also lead to the loss of some hidden information in the training samples, and the accuracy of the loss classification model

In addition, cost-sensitivity learning refers to assigning different costs to positive and negative classes, which involves modifying the cost function or objective function of the classification model, increasing the complexity of the classification model and the difficulty of engineering development, and may not guarantee the final classification effect

Images