Prediction model training method and device taking into account prediction accuracy and privacy protection

Abstract

The embodiments of this specification provide a prediction model training method and device that take into account both prediction accuracy and privacy protection. The prediction model includes a feature extraction layer and a prediction layer. When the prediction model is trained, for the first sample containing the first original feature and the first label of the first object, the first object can be extracted from the first original feature. The first actual value of multiple privacy attributes; input the first original feature into the feature extraction layer to obtain the first extracted feature; input the first extracted feature into the prediction layer to obtain the first prediction information; based on the first prediction information and the first The difference between labels determines the first prediction loss; the first extracted features are input into the pre-trained privacy-preserving model to obtain the first predicted value of the first object in terms of multiple privacy attributes, based on the first predicted value and the first actual value The difference between them determines the second prediction loss; the feature extraction layer is updated in the direction of reducing the first prediction loss and increasing the second prediction loss.

Description

technical field [0001] One or more embodiments of this specification relate to the technical field of machine learning, and in particular to a prediction model training method and device that take into account prediction accuracy and privacy protection. Background technique [0002] With the continuous development of computer technology, the application range of training forecasting models by means of machine learning and using the forecasting models for business forecasting is becoming more and more extensive. For example, a predictive model can be trained to classify users based on user characteristics, such as high-risk users or low-risk users. In the application of the recommendation system, the prediction model can recommend products, stores or other information to the user according to the characteristics of the user, so that the user can obtain the required information more easily. The prediction system can also make predictions based on the event characteristics of ...

AI Technical Summary

Problems solved by technology

An accurate prediction system can provide good prediction results and services, but an overly accurate prediction model may bring concerns about privacy leaks to relevant users and companies, and it also increases the risk of malicious actors stealing user accounts and enterprise accounts and combining predictions. Risk of stealing private information by model

Images