Model multi-terminal collaborative training method and medical risk prediction method and device

Abstract

The embodiment of the invention discloses a model multi-terminal collaborative training method and a medical risk prediction method and device. The method comprises the steps: a first sample model issent to a first terminal and a second sample model is sent to a second terminal; an encrypted first to-be-updated parameter sent by the first terminal and an encrypted second to-be-updated parameter sent by the second terminal are received; the encrypted first to-be-updated parameter and the encrypted second to-be-updated parameter are fused into a target parameter according to a preset encrypteddata processing mode, and the target parameter is sent to the first terminal and the second terminal, wherein the target parameter is used for updating the first sample model and the second sample model; the updated first sample model and the updated second sample model are determined as prediction models, wherein the prediction model is used for predicting a medical risk prediction trend matchedwith the target medical data of the target user. By adopting the methods and the device disclosed in the invention, the prediction accuracy of the prediction model can be improved.

Description

technical field [0001] The present application relates to the field of computer technology, and in particular to a model multi-terminal collaborative training method, a medical risk prediction method, a device, a computer device and a storage medium. Background technique [0002] In recent years, data platforms and technologies such as big data, cloud computing, Internet of Things, deep learning, and artificial intelligence have made great progress in the fields of the Internet and finance, and are widely used in fields such as user portraits, product recommendations, and business operations. However, these The recommendation model or prediction model requires sufficient sample data as support. [0003] However, the sample data often belong to different institutional departments. In order to protect the privacy and security of the sample data stored by the institutional department, they will not disclose the sample data to the public, and will only conduct closed model train...

AI Technical Summary

Problems solved by technology

[0004] The number of sample data stored in a single institution is limited, and the number of sample data participating in model training will directly determine the performance of the model, so the prediction accuracy of the prediction model trained based on a small amount of sample data is low

Images