Method for discovering causality from data, electronic device and storage medium

Patent metadata

Field	Value
Publication number	US-11947552-B2
Application number	US-202217947659-A
Country	US
Kind code	B2
Filing date	Sep 19, 2022
Priority date	Feb 25, 2022
Publication date	Apr 2, 2024
Grant date	Apr 2, 2024

Abstract

A method for discovering causality from data includes acquiring to-be-processed data, and obtaining a covariance matrix of the to-be-processed data; determining a first target column in the covariance matrix, taking the number of columns of the first target column as a first place in a rearrangement sequence, and obtaining a first upper triangular matrix according to the first target column; determining a position of the number of columns of the covariance matrix other than the first target column except the first place in the rearrangement sequence according to the first target column and the first upper triangular matrix, and obtaining an upper triangular matrix in each position determination; obtaining an adjacency matrix according to an upper triangular matrix and a rearrangement sequence obtained in final position determination; and generating directed acyclic graph (DAG) by using the adjacency matrix, and taking the DAG as causality discovery result of the to-be-processed data.

First claim

What is claimed is:

1. A computer-implemented method for discovering causality from data, comprising:
acquiring to-be-processed data which is body data corresponding to different users, and obtaining a covariance matrix of the to-be-processed data;
determining a first target column in the covariance matrix, taking the number of columns of the first target column as a first place in a rearrangement sequence, and obtaining a first upper triangular matrix according to the first target column;
determining a position of the number of columns of the covariance matrix other than the first target column except the first place in the rearrangement sequence according to the first target column and the first upper triangular matrix, and obtaining an upper triangular matrix in each position determination;
obtaining an adjacency matrix according to an upper triangular matrix and a rearrangement sequence obtained in final position determination; and
generating a directed acyclic graph (DAG) by using the adjacency matrix, and taking the DAG as a causality discovery result of the to-be-processed data.

2. The method of claim 1 , wherein the determining a first target column in the covariance matrix comprises:
determining a minimum value located on a diagonal line of the covariance matrix; and
taking a column where the minimum value is located as the first target column in the covariance matrix.

3. The method of claim 1 , wherein the obtaining a first upper triangular matrix according to the first target column comprises:
performing calculation according to the number of samples in the to-be-processed data, the first target column in the covariance matrix, and a first preset parameter, and taking a calculation result as the first upper triangular matrix.

4. The method of claim 2 , wherein the obtaining a first upper triangular matrix according to the first target column comprises:
performing calculation according to the number of samples in the to-be-processed data, the first target column in the covariance matrix, and a first preset parameter, and taking a calculation result as the first upper triangular matrix.

5. The method of claim 1 , wherein the determining a position of the number of columns of the covariance matrix other than the first target column except the first place in the rearrangement sequence according to the first target column and the first upper triangular matrix, and obtaining an upper triangular matrix in each position determination comprises:
taking, for a second place in the rearrangement sequence, columns in the covariance matrix other than the first target column as candidate second target columns;
obtaining a plurality of target calculation results according to the first upper triangular matrix and the candidate second target columns, and taking the candidate second target column corresponding to the target calculation result in the plurality of target calculation results that satisfies a preset requirement as a second target column;
taking the number of columns of the second target column in the covariance matrix as the second place in the rearrangement sequence, and obtaining a second upper triangular matrix according to the first upper triangular matrix and the second target column;
taking, for a third place in the rearrangement sequence, columns in the covariance matrix other than the first target column and the second target column as candidate third target columns;
obtaining a plurality of target calculation results according to the second upper triangular matrix and the candidate third target columns, and taking the candidate third target column corresponding to the target calculation result in the plurality of target calculation results that satisfies a preset requirement as a third target column;
taking the number of columns of the third target column in the covariance matrix as the third place in the rearrangement sequence, and obtaining a third upper triangular matrix according to the second upper triangular matrix and the third target column; and
continuously proceeding until the last place in the rearrangement sequence is determined.

6. The method of claim 5 , wherein obtaining a target calculation result according to a previous upper triangular matrix and a current candidate target column comprises:
obtaining a first calculation result according to the previous upper triangular matrix, a previous target column in the covariance matrix, and the current candidate target column in the covariance matrix; and
obtaining the target calculation result according to the first calculation result.

7. The method of claim 6 , wherein the obtaining the target calculation result according to the first calculation result comprises:
obtaining a second calculation result according to the current candidate target column in the covariance matrix, the first calculation result, and the first preset parameter; and
taking at least one of the first calculation result and the second calculation result as the target calculation result.

8. The method of claim 1 , wherein the obtaining an adjacency matrix according to an upper triangular matrix and a rearrangement sequence obtained in final position determination comprises:
removing diagonal elements from the upper triangular matrix obtained in the final position determination to obtain a strictly upper triangular matrix; and
restoring an order of columns in the strictly upper triangular matrix to an initial order according to the rearrangement sequence obtained in the final position determination, and obtaining the adjacency matrix.

9. The method of claim 8 , wherein the removing diagonal elements from the upper triangular matrix obtained in the final position determination to obtain a strictly upper triangular matrix comprises:
filtering, by using a preset threshold, elements in the upper triangular matrix with the diagonal elements removed; and
taking an upper triangular matrix obtained after the elements are filtered as the strictly upper triangular matrix.

10. The method of claim 2 , wherein the determining a position of the number of columns of the covariance matrix other than the first target column except the first place in the rearrangement sequence according to the first target column and the first upper triangular matrix, and obtaining an upper triangular matrix in each position determination comprises:
taking, for a second place in the rearrangement sequence, columns in the covariance matrix other than the first target column as candidate second target columns;
obtaining a plurality of target calculation results according to the first upper triangular matrix and the candidate second target columns, and taking the candidate second target column corresponding to the target calculation result in the plurality of target calculation results that satisfies a preset requirement as a second target column;
taking the number of columns of the second target column in the covariance matrix as the second place in the rearrangement sequence, and obtaining a second upper triangular matrix according to the first upper triangular matrix and the second target column;
taking, for a third place in the rearrangement sequence, columns in the covariance matrix other than the first target column and the second target column as candidate third target columns;
obtaining a plurality of target calculation results according to the second upper triangular matrix and the candidate third target columns, and taking the candidate third target column corresponding to the target calculation result in the plurality of target calculation results that satisfies a preset requirement as a third target column;
taking the number of columns of the third target column in

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Li Xu
• Cai Yunfeng
• Sun Mingming
• Li Ping

Classifications

• G06F16/2465 Primary

  Query processing support for facilitating data mining operations in structured databases · CPC title

• G06F16/2237 Primary

  Vectors, bitmaps or matrices · CPC title

• G06N5/02 Primary

  Knowledge representation; Symbolic representation · CPC title

• G06F16/2455

  Query execution · CPC title

• G06F16/248

  Presentation of query results · CPC title