An unbalanced classification method based on deep reinforcement learning

Abstract

The invention discloses an unbalanced classification method based on deep reinforcement learning, which comprises the following steps: 1) constructing classification tasks and interaction rules of anagent; 2) constructing the action space of the agent; 3) constructing an external environment; 4) constructing a depth neural network model; 5) training the unbalanced classification model based on depth reinforcement learning, that is to say, learning a Q function with the deep neural network model. The present invention applies deep reinforcement learning to the problem of unbalanced classification in supervised learning, to improve the role of minority sample features in classification modeling by giving higher rewards and punishments to minority sample classification actions through rewardfunction. The method makes the agent learn the correct classification strategy in different types of data environment and data of different unbalanced degree, and is applicable to unbalanced binary classification and multi-classification problems. Therefore, the method has practical application values and is worth popularizing.

Description

technical field [0001] The invention relates to the technical field of deep learning, enhanced learning, and unbalanced classification in machine learning, in particular to an unbalanced classification method based on deep enhanced learning. Background technique [0002] For the unbalanced classification problem, the usual solution is to improve it from the perspective of data layer and algorithm layer. However, the improvement method of the data layer will change the original distribution characteristics of the training data, and the resampled data often cannot reflect the distribution characteristics of the real data. For example, the simple copying or interpolation of the minority class samples by the upsampling method will cause the training model to overfit the minority class samples, while the downsampling method will cause the loss of data information due to the deletion of the majority class samples. There are also problems with the method at the algorithm level. Th...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to overcome the deficiencies in the prior art, and proposes an effective, scientific and reasonable imbalanced classification method based on deep reinforcement learning, using deep reinforcement learning to solve the imbalanced classification problem in supervised learning, Give a higher reward and punishment value to the minority class sample classification action through the reward function to improve the role of minority class sample features in classification modeling, so that the agent can learn correct classification in different types of data environments and data with different degrees of imbalance. Strategy

Images