Unbalanced sample classification method and device

Abstract

The embodiment of the invention provides an unbalanced sample classification method and device, and the method comprises the steps: obtaining unbalanced sample data which comprises sample data and feature data; calculating a sample contribution rate by utilizing the sample data and the feature data; screening out sample data within a preset sample contribution threshold according to the sample contribution rate, and determining the sample data as target sample data; and inputting the target sample data into a sample classification model, and performing calculation by using an optimization classification algorithm to obtain a sample classification result. Two variables of a feature value contribution rate and a feature contribution degree are used to eliminate the characteristics and samples with low classification contribution degree are eliminated, the processing of unbalanced sample data is effectively reduced, the efficient classification is realized by using a machine learning algorithm and effective characteristics or samples on the basis, and the classification efficiency is improved on the premise of ensuring the classification quality.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to a method for classifying unbalanced samples and a device for classifying unbalanced samples. Background technique [0002] Various classification problems are often encountered in real life, such as identifying high-quality customers among many loan applicants, insurance companies judging car insurance levels based on vehicles and car owners, and food grading based on food information samples. When the various samples of the classification problem are relatively balanced, it is easy to get very accurate results. However, in the case of very large differences in the proportions of various samples, the sample ratio reaches 1:100, which is called sample imbalance. It is a big challenge to get a more ideal classification effect. [0003] When dealing with the problem of unbalanced samples, it is currently mainly solved through data sampling processing and algorithm adjustme...

AI Technical Summary

Problems solved by technology

[0004] When traditional algorithms are applied to the classification problem of unbalanced samples, there are problems such as complex calculation methods and low classification accuracy.

Images