Classification model prediction result processing method and device, equipment and storage medium

Abstract

The invention discloses a classification model prediction result processing method, apparatus and device, and a storage medium, and relates to artificial intelligence, machine learning and big data technologies in the computer technology. According to the specific implementation scheme, ciphertext calculation is carried out based on a classification prediction result of a ciphertext, category label data of the ciphertext and statistical information, index information of the ciphertext of the classification prediction result is obtained, and a calculation participant does not need to obtain plaintexts of the classification prediction result, the category label data and the statistical information; based on the classification prediction result, the category label data and the statistical information of the ciphertext, the calculation of the index information can be completed by performing ciphertext calculation, the classification prediction result, the category label data and the statistical information of the ciphertext cannot be decrypted in the calculation process of the index information, and the data cannot be leaked. On the premise of ensuring that a classification predictionresult, classification label data and statistical information are not leaked, evaluation index information of a model can be obtained through calculation, and the data safety is improved.

Description

technical field [0001] The present application relates to artificial intelligence, machine learning, and big data technologies in computer technology, and in particular to a processing method, device, device, and storage medium for classification model prediction results. Background technique [0002] Machine learning is widely used in many fields such as finance and medical care. The reason why machine learning can achieve good results in various fields is not only the rapid development of related technologies and the rapid improvement of hardware computing capabilities, but also the explosive growth of data. However, centralized collection of user data for model training also has great privacy and security risks. In order to perform joint analysis / modeling of multi-institutional data while meeting user privacy protection, data security, and related regulatory requirements, federated learning technology based on secure multi-party computing technology emerged as the times r...

AI Technical Summary

Problems solved by technology

[0004] In the prior art, in the process of federated learning, usually based on the plaintext of the model prediction result combined with the test set label, the evaluation index is determined directly through the plaintext calculation, which leaks the prediction result of the model on the test set. If the model index calculation party is not the test set set labels, but also leaks the labels of the test set

Images