Granted Patent B2

US 12,591,783 · App. 18/072,622 · Granted Mar 31, 2026

Inventors: Litao Hong (Shenzhen, CN); Yongyao Chao (Shenzhen, CN)

Assignee: TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED

G06N3/091

Abstract

Claims (77)

1 . A method for constructing a recommendation model performed by an electronic device, the method comprising:

aggregating a plurality of feature tables corresponding to each application scenario in a recommendation project to obtain an aggregated feature table, the recommendation project comprising a plurality of application scenarios in a one-to-one correspondence with a plurality of recommendation indicators of an information item and each application scenario having a recommendation model being configured to predict a corresponding recommendation indicator;

receiving a corresponding user feature and item feature from the aggregated feature table based on a user identifier and an item identifier comprised in a sample data table, and stitching the user feature and the item feature with the sample data table, to form a training sample set;

performing primary key encoding on a feature identifier of each training sample of the training sample set, to obtain a primary key encoded value of the feature identifier;

performing secondary key encoding on the feature identifier of each training sample, to obtain a secondary key encoded value of the feature identifier;

stitching the primary key encoded value and the secondary key encoded value, to obtain an index encoded value of the feature identifier;

updating the feature identifier of the training sample to the index encoded value, to obtain an updated training sample set;

training the recommendation model of the application scenario based on the updated training sample set, the trained recommendation model being capable of fitting a user feature and an item feature in the updated training sample set; and

recommending information items to at least one user using the trained recommendation model.

2 . The method according to claim 1 , wherein the aggregating a plurality of feature tables corresponding to each application scenario in a recommendation project to obtain an aggregated feature table comprises:

aggregating and deduplicating at least a part of features of the plurality of feature tables corresponding to the application scenario, to obtain an aggregated feature table of the application scenario; and

combining feature identifiers in the aggregated feature table, to obtain a feature metadata table of the application scenario.

3 . The method according to claim 2 , wherein the aggregating at least a part of features of the plurality of feature tables corresponding to the application scenario comprises:

aggregating all features of the plurality of feature tables corresponding to the application scenario.

4 . The method according to claim 2 , wherein the aggregating at least a part of features of the plurality of feature tables corresponding to the application scenario comprises:

determining, from the plurality of feature tables corresponding to each application scenario in the recommendation project, features publicly used by a plurality of training algorithms for training the recommendation model of the application scenario; and

aggregating the publicly used features, to obtain the aggregated feature table of the application scenario.

5 . The method according to claim 1 , wherein when a newly added feature table exists in the application scenario, the method further comprises:

stitching the newly added feature table with the aggregated feature table of the application scenario, to obtain a new aggregated feature table; and

incrementally updating a cache space based on the new aggregated feature table.

6 . The method according to claim 1 , wherein the aggregating a plurality of feature tables corresponding to each application scenario in a recommendation project comprises:

aggregating newly added feature tables corresponding to new periods when each new period of each application scenario arrives, to obtain an aggregated feature table of the new period; and

stitching aggregated feature tables of the new periods, to obtain the aggregated feature table of the application scenario.

7 . The method according to claim 1 , wherein the receiving a corresponding user feature and item feature from the aggregated feature table based on a user identifier and an item identifier comprised in a sample data table, and stitching the user feature and the item feature with the sample data table comprises:

when each new period of each application scenario arrives:

reading a corresponding user feature and item feature from an aggregated feature table of the new period in a cache space based on a user identifier and an item identifier comprised in a sample data table of the new period; and

stitching the user feature and the item feature with the sample data table of the new period, to obtain cache features of the new period.

8 . The method according to claim 1 , further comprising:

mapping feature identifiers in the obtained aggregated feature table, to obtain integer values of the feature identifiers;

updating the feature identifiers in the aggregated feature table to the integer values, to obtain a compressed aggregated feature table; and

transmitting the compressed aggregated feature table to a cache space.

9 . An electronic device, comprising:

a memory, configured to store executable instructions; and

a processor, configured to perform, when executing the executable instructions stored in the memory, a method for constructing a recommendation model, the method including:

aggregating a plurality of feature tables corresponding to each application scenario in a recommendation project to obtain an aggregated feature table, the recommendation project comprising a plurality of application scenarios in a one-to-one correspondence with a plurality of recommendation indicators of an information item and each application scenario having a recommendation model being configured to predict a corresponding recommendation indicator;

receiving a corresponding user feature and item feature from the aggregated feature table based on a user identifier and an item identifier comprised in a sample data table, and stitching the user feature and the item feature with the sample data table, to form a training sample set;

performing primary key encoding on a feature identifier of each training sample of the training sample set, to obtain a primary key encoded value of the feature identifier;

performing secondary key encoding on the feature identifier of each training sample, to obtain a secondary key encoded value of the feature identifier;

stitching the primary key encoded value and the secondary key encoded value, to obtain an index encoded value of the feature identifier;

updating the feature identifier of the training sample to the index encoded value, to obtain an updated training sample set;

training the recommendation model of the application scenario based on the updated training sample set, the trained recommendation model being capable of fitting a user feature and an item feature in the updated training sample set; and

recommending information items to at least one user using the trained recommendation model.

10 . The electronic device according to claim 9 , wherein the aggregating a plurality of feature tables corresponding to each application scenario in a recommendation project to obtain an aggregated feature table comprises:

aggregating and deduplicating at least a part of features of the plurality of feature tables corresponding to the application scenario, to obtain an aggregated feature table of the application scenario; and

combining feature identifiers in the aggregated feature table, to obtain a feature metadata table of the application scenario.

11 . The electronic device according to claim 9 , wherein when a newly added feature table exists in the application scenario, the method further comprises:

stitching the newly added feature table with the aggregated feature table of the application scenario, to obtain a new aggregated feature table; and

incrementally updating a cache space based on the new aggregated feature table.

12 . The electronic device according to claim 9 , wherein the aggregating a plurality of feature tables corresponding to each application scenario in a recommendation project comprises:

aggregating newly added feature tables corresponding to new periods when each new period of each application scenario arrives, to obtain an aggregated feature table of the new period; and

stitching aggregated feature tables of the new periods, to obtain the aggregated feature table of the application scenario.

13 . The electronic device according to claim 9 , wherein the receiving a corresponding user feature and item feature from the aggregated feature table based on a user identifier and an item identifier comprised in a sample data table, stitching the user feature and the item feature with the sample data table comprises:

when each new period of each application scenario arrives:

reading a corresponding user feature and item feature from an aggregated feature table of the new period in a cache space based on a user identifier and an item identifier comprised in a sample data table of the new period; and

stitching the user feature and the item feature with the sample data table of the new period, to obtain cache features of the new period.

14 . The electronic device according to claim 9 , wherein the method further comprises:

mapping feature identifiers in the obtained aggregated feature table, to obtain integer values of the feature identifiers;

updating the feature identifiers in the aggregated feature table to the integer values, to obtain a compressed aggregated feature table; and

transmitting the compressed aggregated feature table to a cache space.

15 . A non-transitory computer-readable storage medium, storing executable instructions that, when executed by a processor of a computer device, implements a method for constructing a recommendation model, the method including:

aggregating a plurality of feature tables corresponding to each application scenario in a recommendation project to obtain an aggregated feature table, the recommendation project comprising a plurality of application scenarios in a one-to-one correspondence with a plurality of recommendation indicators of an information item and each application scenario having a recommendation model being configured to predict a corresponding recommendation indicator;

receiving a corresponding user feature and item feature from the aggregated feature table based on a user identifier and an item identifier comprised in a sample data table, and stitching the user feature and the item feature with the sample data table, to form a training sample set;

performing primary key encoding on a feature identifier of each training sample of the training sample set, to obtain a primary key encoded value of the feature identifier;

performing secondary key encoding on the feature identifier of each training sample, to obtain a secondary key encoded value of the feature identifier;

stitching the primary key encoded value and the secondary key encoded value, to obtain an index encoded value of the feature identifier;

updating the feature identifier of the training sample to the index encoded value, to obtain an updated training sample set;

training the recommendation model of the application scenario based on the updated training sample set, the trained recommendation model being capable of fitting a user feature and an item feature in the updated training sample set; and

recommending information items to at least one user using the trained recommendation model.

16 . The non-transitory computer-readable storage medium according to claim 15 , wherein the aggregating a plurality of feature tables corresponding to each application scenario in a recommendation project to obtain an aggregated feature table comprises:

aggregating and deduplicating at least a part of features of the plurality of feature tables corresponding to the application scenario, to obtain an aggregated feature table of the application scenario; and

combining feature identifiers in the aggregated feature table, to obtain a feature metadata table of the application scenario.

17 . The non-transitory computer-readable storage medium according to claim 15 , wherein when a newly added feature table exists in the application scenario, the method further comprises:

stitching the newly added feature table with the aggregated feature table of the application scenario, to obtain a new aggregated feature table; and

incrementally updating a cache space based on the new aggregated feature table.

18 . The non-transitory computer-readable storage medium according to claim 15 , wherein the aggregating a plurality of feature tables corresponding to each application scenario in a recommendation project comprises:

aggregating newly added feature tables corresponding to new periods when each new period of each application scenario arrives, to obtain an aggregated feature table of the new period; and

stitching aggregated feature tables of the new periods, to obtain the aggregated feature table of the application scenario.

Assignments (1)

Priority Claims (1)

Continuity (2)

References Cited (16)