Social content risk identification

1. A computer-implemented method for social content risk identification, comprising:

obtaining, by a computer, social content data comprising at least one of social behavior records or social message records;

extracting, by the computer, original features of the social content data;

for each of the original features, generating, by the computer and using a tree structured machine learning model, dimension-extended features that include a set of extended sub-features corresponding to different dimensions of the original feature;

classifying, by the computer and using a deep machine learning model, that the social content data is risky social content data based on the dimension-extended features; and

transmitting, by the computer, a result of classifying that the social content data is risky social content data to a risk control platform.

2. The computer-implemented method of claim 1 , wherein the tree structured machine learning model and the deep machine learning model have been pre-trained by using black samples that are risky and white samples that are not risky.

3. The computer-implemented method of claim 1 , wherein extracting the original features of the social content data comprises:

performing, by the computer, data cleaning on the social content data; and

extracting, by the computer based on feature engineering, the original features of the at least one of social behavior records or social message records from the social content data after the data cleaning.

4. The computer-implemented method of claim 1 , wherein generating dimension-extended features using the tree structured machine learning model comprises:

inputting, by the computer, each original feature into the tree structured machine learning model that outputs prediction data of the original feature in a plurality of leaf nodes; and

generating, by the computer, the dimension-extended features of the original feature based on the prediction data of the plurality of leaf nodes.

5. The computer-implemented method of claim 1 , wherein after classifying that the social content data is the risky social content data, the method further comprises:

based on timed scheduling, periodically obtaining, by the computer, updated social content data and providing a risk identification result of the updated social content data to the risk control platform.

6. The computer-implemented method of claim 1 , wherein the tree structured machine learning model comprises a gradient boosting decision tree (GBDT).

7. The computer-implemented method of claim 1 , wherein the deep machine learning model comprises a deep neural network (DNN).

8. The computer-implemented method of claim 1 , wherein the risky social content data comprises: misconducts or misstatements related to a specified field.

9. A non-transitory, computer-readable medium storing one or more instructions executable by a computer system to perform operations comprising:

obtaining, by the computer system, social content data comprising at least one of social behavior records or social message records;

extracting, by the computer system, original features of the social content data;

for each of the original features, generating, by the computer system and using a tree structured machine learning model, dimension-extended features that include a set of extended sub-features corresponding to different dimensions of the original feature;

classifying, by the computer system and using a deep machine learning model, that the social content data is risky social content data based on the dimension-extended features; and

transmitting, by the computer system, a result of classifying that the social content data is risky social content data to a risk control platform.

10. The non-transitory, computer-readable medium of claim 9 , wherein the tree structured machine learning model and the deep machine learning model have been pre-trained by using black samples that are risky and white samples that are not risky.

11. The non-transitory, computer-readable medium of claim 9 , wherein extracting the original features of the social content data comprises:

performing, by the computer system, data cleaning on the social content data; and

extracting, by the computer system based on feature engineering, the original features of the at least one of social behavior records or social message records from the social content data after the data cleaning.

12. The non-transitory, computer-readable medium of claim 9 , wherein generating dimension-extended features using the tree structured machine learning model comprises:

inputting, by the computer system, each original feature into the tree structured machine learning model that outputs prediction data of the original feature in a plurality of leaf nodes; and

generating, by the computer system, the dimension-extended features of the original feature based on the prediction data of the plurality of leaf nodes.

13. The non-transitory, computer-readable medium of claim 9 , wherein after classifying that the social content data is the risky social content data, the operations further comprise:

based on timed scheduling, periodically obtaining, by the computer system, updated social content data and providing a risk identification result of the updated social content data to the risk control platform.

14. The non-transitory, computer-readable medium of claim 9 , wherein the tree structured machine learning model comprises a gradient boosting decision tree (GBDT).

15. The non-transitory, computer-readable medium of claim 9 , wherein the deep machine learning model comprises a deep neural network (DNN).

16. The non-transitory, computer-readable medium of claim 9 , wherein the risky social content data comprises: misconducts or misstatements related to a specified field.

17. A computer-implemented system, comprising:

one or more computers; and

one or more computer memory devices interoperably coupled with the one or more computers and having tangible, non-transitory, machine-readable media storing one or more instructions that, when executed by the one or more computers, perform operations comprising:

obtaining, by the one or more computers, social content data comprising at least one of social behavior records or social message records;

extracting, by the one or more computers, original features of the social content data;

for each of the original features, generating, by the one or more computers and using a tree structured machine learning model, dimension-extended features that include a set of extended sub-features corresponding to different dimensions of the original feature;

classifying, by the one or more computers and using a deep machine learning model, that the social content data is risky social content data based on the dimension-extended features; and

transmitting, by the one or more computers, a result of classifying that the social content data is risky social content data to a risk control platform.

18. The computer-implemented system of claim 17 , wherein the tree structured machine learning model and the deep machine learning model have been pre-trained by using black samples that are risky and white samples that are not risky.

19. The computer-implemented system of claim 17 , wherein extracting the original features of the social content data comprises:

performing, by the one or more computers, data cleaning on the social content data; and

extracting, by the one or more computers based on feature engineering, the original features of the at least one of social behavior records or social message records from the social content data after the data cleaning.

20. The computer-implemented system of claim 17 , wherein generating dimension-extended features using the tree structured machine learning model comprises:

inputting, by the one or more computers, each original feature into the tree structured machine learning model that outputs prediction data of the original feature in a plurality of leaf nodes; and

generating, by the one or more computers, the dimension-extended features of the original feature based on the prediction data of the plurality of leaf nodes.

21. The computer-implemented system of claim 17 , wherein after classifying that the social content data is the risky social content data, the operations further comprise:

based on timed scheduling, periodically obtaining, by the one or more computers, updated social content data and providing a risk identification result of the updated social content data to the risk control platform.

22. The computer-implemented system of claim 17 , wherein the tree structured machine learning model comprises a gradient boosting decision tree (GBDT).

23. The computer-implemented system of claim 17 , wherein the deep machine learning model comprises a deep neural network (DNN).

24. The computer-implemented system of claim 17 , wherein the risky social content data comprises: misconducts or misstatements related to a specified field.