Federated learning-based feature completion method, equipment and storage medium

Abstract

The invention discloses a federated learning-based feature completion method and a device thereof, equipment and a storage medium, and the method comprises the steps: obtaining to-be-completed sample data, and inputting the to-be-completed sample data into a preset target feature completion model; wherein the preset target feature completion model is obtained by performing iterative training on a preset initial completion model by executing a first preset longitudinal federated learning process based on non-missing sample feature data with a label; and performing completion processing on the to-be-completed sample data based on the feature completion model to obtain target completion data so as to use the target completion data to perform training of a machine learning model. According to the method and the device, the preset target feature completion model is jointly established by utilizing different features of the plurality of participant devices, so that the condition that the prediction accuracy of the model obtained by training is low when the number of samples is relatively small or the sample features are relatively homogeneous is avoided, namely, the prediction accuracy of the feature prediction model is improved.

Description

technical field [0001] The present application relates to the field of artificial intelligence, and in particular to a federated learning-based feature completion method, device, device, and storage medium. Background technique [0002] With the continuous development of financial technology, especially Internet technology and finance, more and more technologies (such as distributed, blockchain, artificial intelligence, etc.) Requirements, such as the completion of missing features in the financial industry also have higher requirements. [0003] With the continuous development of computer software and artificial intelligence, neural network models are more and more widely used. However, neural network models with superior performance are usually constructed based on training samples with high feature richness. In real application scenarios, feature The lack of value is normal. For example, in the field of financial risk control, institutions engaged in financial risk contr...

AI Technical Summary

Problems solved by technology

[0004] The main purpose of this application is to provide a feature completion method, device, device, and storage medium based on federated learning, aiming to solve the problems in the prior art by using average values, medians, and mode and other feature value completion methods. To fill the eigenvalues ​​of the participants' equipment vacancies, it is difficult to obtain a prediction model with high prediction accuracy.

Images