Method and device for training event prediction model

A technology for predicting models and events, applied in computing models, neural learning methods, biological neural network models, etc., can solve problems such as limited evaluation accuracy and single evaluation methods

Active Publication Date: 2020-12-18
ALIPAY (HANGZHOU) INFORMATION TECH CO LTD
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the current evaluation method is relatively simple, resulting in very limited evaluation accuracy

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Method and device for training event prediction model
  • Method and device for training event prediction model
  • Method and device for training event prediction model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment approach

[0064] According to an implementation manner, the above-mentioned first combination further includes an N-order inter-vector combination operation involving multiplication of N coded vectors, where N>=2.

[0065] It can be understood that the combination of feature vectors in conventional neural networks generally adopts a linear combination method. However, when an event contains multiple attribute information, sometimes the attribute information is not completely independent, but there is a certain dependency or correlation, and a simple linear combination is not enough to discover and process such a correlation. Therefore, in one embodiment, referring to the framework of the FM factorization machine, a high-order inter-vector combination operation is introduced in the first embedding layer.

[0066] The N-order inter-vector combination operation involves the multiplication operation of N coded vectors, so that the association relationship between the N coded vectors can be ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The embodiment of this specification provides a method for training an event prediction model. This method can be applied to the scenario of transfer learning. By setting up a neutral server, the data isolation and privacy security protection of the source domain participant and the target domain participant are realized. Among them, the source domain participant deploys the source domain feature extractor, the target domain participant deploys the target domain feature extractor, and the neutral server deploys the common part of the model in the event prediction model, including the shared feature extractor, graph neural network and classification network. For any participant, it uses the feature extractor of the local domain to extract the features of the samples in the domain to obtain the feature representation of the domain, and then uses the current parameters of the common part of the model obtained from the server to perform the feature representation of the domain. processing to obtain the corresponding event classification results, and then update the model based on the event classification results and local samples, and upload the update results of the common part of the model to the server, so that the server can perform centralized updates.

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 training an event prediction model using machine learning. Background technique [0002] In many scenarios, user operation behavior or operation events need to be analyzed and processed. For example, in order to identify high-risk operations that may threaten network security or user information security, such as account theft, traffic attacks, fraudulent transactions, etc., the risk level of user operations can be assessed for risk prevention and control. [0003] In order to assess the risk degree of an operation behavior, it can be analyzed based on the characteristics of the operation behavior itself. Further, the user's behavior sequence can also be considered more comprehensively. Behavior sequence is the occurrence process of a series of clicks, visits, purchases and other events generated by users in daily opera...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Patents(China)
IPC IPC(8): G06N3/08G06N20/00G06K9/62
CPCG06N3/08G06N20/00G06F18/241G06F18/214
Inventor 宋博文顾曦陈帅
Owner ALIPAY (HANGZHOU) INFORMATION TECH CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products