Method and device for jointly determining object feature correlation in private data by multiple parties

Abstract

The embodiment of the invention provides a method and device for jointly determining object feature correlation in private data by multiple parties. The privacy data are distributed in a plurality of holders, and the first holder performs zero mean on feature values of multiple features in the first original matrix to obtain a first center matrix; and a first fragment matrix of the covariance matrix is determined based on the first central matrix and respective central matrixes of other owners by using multi-party security calculation. For the ith feature in the first holding party and the jth feature in the second holding party, the first holding party obtains data from the local covariance matrix fragment and the locally stored feature data, and determines a first correlation coefficient fragment between the ith feature and the jth feature based on the local feature data of the second holding party by using multi-party security calculation.

Description

technical field [0001] One or more embodiments of this specification relate to the technical field of data processing, and in particular to a method and device for multiple parties to jointly determine the correlation of object features in private data. Background technique [0002] In the process of constructing linear models in the real world, some sample features will be related to other sample features, that is, multicollinearity of features or variables, that is, there is a high correlation between sample features. When this situation is serious, it will increase the variance of the regression coefficient in the model, making the result of the regression model unstable. In the multi-platform joint modeling scenario, due to the joint modeling of multiple parties, the sample data used are generated in similar scenarios, which will inevitably produce collinear features with different names, so the correlation between features before building a linear model It is very nece...

AI Technical Summary

Problems solved by technology

However, the sample data in each platform often has strong privacy protection requirements, and will not be transmitted in plaintext or simple ciphertext.

Images