A prediction method and device based on tree model

Abstract

The embodiment of this specification provides a privacy-protecting tree model construction method and device and a prediction method and device based on the tree model. The construction method includes: obtaining M groups of split results from the respective devices of at least two data parties, M Group splitting results correspond to M features respectively; record the corresponding data cubes of M group splitting results; calculate the splitting gain of each splitting result based on the respective label values ​​of N samples; obtain the splitting result with the largest splitting gain as the optimal splitting result; in the case where the splitting gain of the optimal splitting result is positive, determine the data cube corresponding to the optimal splitting result; in the case that the corresponding data cube is the second data cube, send the optimal splitting result to the second The device of the second data cube, and record the corresponding relationship between the first node and the second data cube; mark the first node to indicate that there is no node data of the first node locally, and update the tree structure of the first tree accordingly .

Description

technical field [0001] The embodiments of this specification relate to the technical field of machine learning, and more specifically, relate to a method and device for constructing a tree model, and a method and device for predicting based on a tree model. Background technique [0002] Machine learning that can protect data security is currently a technical issue of widespread concern. That is, multiple data parties have their own data, and under the condition of ensuring the data security of all parties, all parties cooperate to train the machine learning model for multiple parties to use. The linear regression / logistic regression tree model is a regression / classification model widely used in the industry. In the case of use, we often face the following problems: multiple data parties have their own data, and they want to use each other's data to unify the modeling tree model, such as Xgboos t or GBDT model, but they do not want to provide their own data to each other . ...

AI Technical Summary

Problems solved by technology

In the case of use, the following problems are often faced: multiple data parties have their own data, and they want to use each other's data to unify the modeling tree model, such as Xgboos t or GBDT model, but they do not want to provide their own data to each other

Images