Binning data processing method, device and equipment, storage medium and program product

Abstract

The invention discloses a binning data processing method, device, equipment, a storage medium and a program product, and the method comprises the steps: obtaining a plurality of data IDs sent by a target variable provider, an encrypted target variable corresponding to each data ID, and an encrypted opposite variable, and calculating an encrypted positive sample proportion and an encrypted negative sample proportion corresponding to each sub-box, multiplying the encrypted positive sample proportion and the encrypted negative sample proportion by corresponding random numbers, and sending multiplication results to a target variable provider, and obtaining an intermediate result corresponding to each sub-box determined by the target variable provider according to the multiplication result, and sending the encrypted information value and / or a result obtained by adding the random number to the evidence weight to the target variable provider according to the intermediate results corresponding to the plurality of sub-boxes and the corresponding random number. According to the method, the security of solving the information value and the evidence weight can be improved.

Description

Technical field [0001] The present invention relates to data processing technology, and in particular, to a method of processing data binning, apparatus, equipment, storage medium, and a program product. Background technique [0002] Learning can combine multiple federal agencies, while meeting user privacy and data security, machine learning model. [0003] During the federal study, WOE (Weight of Evidence, weight of evidence) and IV (InformationValue, the value of information) is a very important indicator, can be used to assess the predictive power characteristic variables. In practice, one of the target holders of variable, other variables held features, both of the interaction may be implemented WOE calculated and IV, thereby completing the filter characteristic variables and so on. [0004] However, if the objective variable target variable will hold one special encoding, it is possible characteristic variable taking information from the other, leading to leakage of data, s...

AI Technical Summary

Problems solved by technology

[0004] However, if the party holding the target variable encodes the target variable in a special way, it is possible to extract the information of the characteristic variable from the other party, resulting in data leakage. Therefore, the existing calculation process of WOE and IV has poor security.

Images