Method and device for constructing and predicting isolated forest model based on federated learning

Abstract

The embodiment of this specification provides a method and device for constructing an isolated forest model based on federated learning. The method includes: acquiring multiple sample identifiers corresponding to the first node, the multiple sample identifiers corresponding to multiple samples, each The sample includes eigenvalues ​​of m features; a feature identifier is randomly selected from the m feature identifiers; in the case that the selected feature identifier is the first feature identifier, based on the local storage of the first feature identifier and the first data party Correspondence, sending the first node’s identifier, multiple sample identifiers, and the first feature identifier to the first data party; recording the correspondence between the first node and the first data party; receiving from the first data party and the first node’s The information corresponding to the two sub-nodes, so as to protect the private data of each data party while building an isolated forest model for business processing.

Description

technical field [0001] The embodiment of this description relates to the field of machine learning technology, and more specifically, to a method and device for constructing an isolated forest model based on federated learning, and a method and device for predicting abnormality of objects through the isolated forest model based on federated learning. Background technique [0002] At present, more and more Internet companies as data owners are beginning to pay attention to data privacy and data security issues. The isolation forest model is an unsupervised learning model that predicts abnormal objects. For example, this model can be used to analyze user behavior to identify abnormal behavior, thereby protecting the security of user funds, such as theft risk prevention and fraud risk prevention and control, etc. However, data modeling in the above scenarios is often carried out under the condition of data fusion (that is, data centralized storage / visibility), which often requi...

AI Technical Summary

Problems solved by technology

However, data modeling in the above scenarios is often carried out under the condition of data fusion (that is, data centralized storage / visibility), which often requires that data from different sources need to be fully exposed to each other to complete the modeling and analysis work. The level of privacy data is very risky

Images