Method and device for multi-party joint dimension reduction processing of private data

Abstract

The embodiment of this specification provides a multi-party joint dimension reduction method and device for private data. In the case of vertical distribution of private data, the first holder performs zero-meanization on the first original matrix to obtain the first center matrix. An N*N-dimensional asymmetric orthogonal matrix is ​​obtained, the asymmetric orthogonal matrix is ​​multiplied by the first center matrix to obtain a first secret matrix, and the first secret matrix is ​​sent to a trusted third party. A trusted third party concatenates each hidden matrix to get the global hidden matrix, multiplies the global hidden matrix with its transposed matrix to get the covariance matrix, solves the eigenvalues ​​of the covariance matrix to get the dimensionality reduction transformation matrix, and calculates the dimensionality reduction transformation matrix After splitting, each split matrix is ​​obtained and sent to the holder. The first holder uses the first splitting matrix to process the first original matrix to obtain the first dimensionality reduction matrix, which is used to perform business forecast analysis on the business object by means of machine learning.

Description

technical field [0001] One or more embodiments of this specification relate to the field of machine learning, and in particular to a method and device for multi-party joint dimensionality reduction processing on private data. Background technique [0002] The data needed for machine learning often involves multiple platforms and fields. For example, in the merchant classification analysis scenario based on machine learning, the electronic payment platform has the transaction flow data of the merchants, the e-commerce platform stores the sales data of the merchants, and the banking institution has the loan data of the merchants. Data often exists in silos. Due to issues such as industry competition, data security, and user privacy, data integration is facing great resistance. How to integrate data scattered across various platforms under the premise of ensuring that data is not leaked has become a challenge. [0003] On the other hand, as the amount of data increases, the d...

AI Technical Summary

Problems solved by technology

Although a large amount of high-dimensional data can enrich the training sample data of machine learning, in fact, these high-dimensional data often have some redundant information

The help of redundant information to the effect of machine learning is very limited, and the resulting high-dimensional feature data may cause "dimension explosion", making it difficult for machine learning models to handle and affecting the training efficiency of the model

Images