Data processing method

Patent metadata

Field	Value
Publication number	US-2024330328-A1
Application number	US-202418741744-A
Country	US
Kind code	A1
Filing date	Jun 12, 2024
Priority date	Jun 13, 2023
Publication date	Oct 3, 2024
Grant date	—

Abstract

A method is provided. The method includes: obtaining an object relationship diagram; for a target object of a plurality of first objects, obtaining at least one meta-path corresponding to the target object in the object relationship diagram; for each meta-path, performing the following operations: determining a plurality of first attention weights of the target object based on inherent attribute data of the target object and inherent attribute data of each of a plurality of second objects on the meta-path; obtaining a second representation vector of the target object based on a first representation vector of the target object and the plurality of first attention weights; and obtaining a target indicator prediction result of the target object based at least on at least one second representation vector of the target object corresponding to the at least one meta-path.

First claim

1 . A data processing method, comprising:
obtaining an object relationship diagram, wherein the object relationship diagram comprises a plurality of first nodes corresponding to a plurality of first objects and is used to represent association relationship information among the plurality of first objects; obtaining, for a target object of the plurality of first objects, at least one meta-path corresponding to the target object in the object relationship diagram, wherein each meta-path of the at least one meta-path is a node path for connecting the target object and an associated object of the target object in the object relationship diagram, and wherein the target object determines a plurality of associated paths with a plurality of second objects other than the target object in the meta-path, respectively; performing, for each meta-path of the at least one meta-path, following operations:
determining, based on inherent attribute data of the target object and inherent attribute data of each of the plurality of second objects on the meta-path, a plurality of first attention weights of the target object relative to the plurality of associated paths respectively; and
obtaining, based on a first representation vector of the target object and the plurality of first attention weights, a second representation vector of the target object that fused association relationship information represented by the plurality of associated paths on the meta-path, wherein the first representation vector is determined based on the inherent attribute data of the target object; and
obtaining, based at least on at least one second representation vector of the target object corresponding to the at least one meta-path, a target indicator prediction result of the target object.

2 . The method according to claim 1 , wherein the plurality of first objects include a plurality of managing subjects and a plurality of managed objects, the target object and each of the plurality of second objects being managed objects, the plurality of associated paths including a plurality of association types, and wherein the plurality of association types include at least two of following association types: direct association between two managed objects, indirect association between two managed objects based on at least one managing subject, indirect association between two managed objects based on at least one managed object, and indirect association between two managed objects based on at least one managing subject and at least one managed object, and wherein
the determining, based on the inherent attribute data of the target object and the inherent attribute data of each of the plurality of second objects on the meta-path, the plurality of first attention weights of the target object relative to the plurality of associated paths respectively comprises: performing, for each association type of the plurality of association types, following operations:
determining, in the plurality of associated paths, at least one associated path corresponding to the association type and at least one second object corresponding to the at least one associated path; and
determining, based on the first representation vector of the target object and third representation vector corresponding to each of the at least one second object, at least one second attention weight of the target object relative to the at least one associated path respectively, corresponding to the association type, wherein the third representation vector is determined based on the inherent attribute data of corresponding second object; and
determining, based on at least one second attention weight corresponding to each association type of the plurality of association types, the plurality of first attention weights.

3 . The method according to claim 2 , wherein the obtaining, based on the first representation vector of the target object and the plurality of first attention weights, the second representation vector of the target object that fused association relationship information represented by the plurality of associated paths on the meta-path comprises:
obtaining, for each association type of the plurality of association types, a fourth representation vector corresponding to the association type based on the first representation vector of the target object and at least one second attention weight corresponding to the association type of the plurality of first attention weights; and concatenating a plurality of fourth representation vectors corresponding to the plurality of association types to obtain the second representation vector.

4 . The method according to claim 3 , wherein the at least one meta-path is plural, and wherein the obtaining, based at least on the at least one second representation vector of the target object corresponding to the at least one meta-path, the target indicator prediction result of the target object comprises:
determining, for each meta-path of the at least one meta-path, a path representation vector of the meta-path based on second representation vector corresponding to each managed object on the meta-path; determining, based on the path representation vector corresponding to each of the at least one meta-path, at least one third attention weight corresponding to the at least one meta-path; obtaining, based on the at least one second representation vector and the at least one third attention weight, a fifth representation vector of the target object; and obtaining, based at least on the fifth representation vector, the target indicator prediction result.

5 . The method according to claim 4 , wherein each managed object of the plurality of managed objects is composed of a plurality of sub-objects in proportion, and wherein the method further comprises:
obtaining tensor data corresponding to the object relationship diagram, wherein the tensor data is used to record proportion data of each sub-object corresponding to each managed object in the object relationship diagram; and performing tensor decomposition on the tensor data to obtain a sixth representation vector corresponding to each managed object in the object relationship diagram, wherein the sixth representation vector is used to represent a component feature of a corresponding managed object; and wherein the obtaining, based at least on the fifth representation vector, the target indicator prediction result comprises:
performing feature fusion on the fifth representation vector and the sixth representation vector of the target object to obtain a seventh representation vector of the target object; and
obtaining, based at least on the seventh representation vector of the target object, the target indicator prediction result.

6 . The method according to the claim 5 , wherein the target indicator prediction result comprises a target indicator prediction value of the target object on a target date, and wherein the method further comprises:
performing, for each first date of at least one first date prior to the target date, following operations:
determining, based on target dynamic attribute data on the first date, a preset number of third objects in the plurality of managed objects, wherein difference between the dynamic attribute data of the third object and the target dynamic attribute data is less than difference between the target dynamic attribute data and the dynamic attribute data of other managed object other than the third object in the plurality of managed objects; and
performing feature fusion on a seventh representation vector corresponding to each third object of the preset number of third objects to obtain a first dynamic feature representation vector corresponding to the first date; and wherein
the obtaining, based at least on the seventh representation vector of the target ob

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Hao Siyuan
• Zhang Le
• Dai Le
• Zhou Jingbo
• Zhang Shengming
• Qin Chuan
• Xiong Hui

Classifications

• G06N3/08

  Learning methods · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06F18/253

  of extracted features · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06F18/213

  Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods · CPC title