Self-adaptive integrated unbalanced data classification method based on Euclidean distance

Abstract

The invention discloses an self-adaptive integrated unbalanced data classification method based on Euclidean distance, which comprises the following steps of: firstly, obtaining a plurality of diversified balance subsets by using a random balance method, then establishing and obtaining a plurality of basic classifiers on each balance subset; and adding a classifier pre-selection algorithm before the dynamic selection algorithm. After a screened basic classifier is obtained, a new dynamic selection algorithm is provided, and by evaluating the condition of the sample classifier in the surrounding area of a to-be-classified sample, the capability is stronger when more minority class samples belong to the correct classification range. And finally, a prediction result obtained by the selected basic classifier by adopting a distance-based adaptive integration rule is output. According to the method, basic classifiers can be established on the generated diversified subsets, meanwhile, a dynamic selection algorithm is provided, the sub-classifier with the highest classification capacity can be selected out, finally, the proposed integration rule can provide a better output result, and finally, the unbalanced data classification precision is effectively improved.

Description

technical field [0001] The invention belongs to the field of artificial intelligence, and in particular relates to an unbalanced data classification method based on Euclidean distance-based self-adaptive integration. Background technique [0002] Imbalanced data refers to the situation in which the number of samples of one class or samples of multiple classes in the training sample is very different from the number of samples of other classes. According to the research report, the class imbalance problem occurs in a variety of real-world domains, such as facial age estimation, detecting oil spills in satellite images, anomaly detection, identifying fraudulent credit card transactions, software defect prediction, and image annotation, etc. Therefore, researchers have attached great importance to the problem of data imbalance and held several symposiums and conferences, such as the Association for the Advancement of Artificial Intelligence (AAAI) 2000, the International Confer...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problem of insufficient diversity of sub-categories in ensemble learning, poor performance basic classifier and ensemble rule design are not considered, this application proposes an unbalanced data classification method based on Euclidean distance-based self-adaptive integration, which improves the unbalanced data classification precision

Images