Methods and apparatuses for building data identification models

1. A method for building a data identification model, comprising:

performing logistic regression training using training samples to obtain a first model, the training samples comprising positive samples and negative samples;

sampling the training samples proportionally to obtain a first training sample set;

identifying the positive samples using the first model, and selecting a second training sample set from positive samples that have identification results after being identified using the first model; and

performing Deep Neural Networks (DNN) training using the first training sample set and the second training sample set to obtain a final data identification model.

2. The method for building a data identification model of claim 1 , wherein prior to sampling the training samples proportionally or performing logistic regression training, the method further comprises:

performing feature engineering preprocessing on the training samples.

3. The method for building a data identification model of claim 2 , wherein prior to performing logistic regression training using the training samples to obtain a first model, the method further comprises:

performing feature screening on the training samples to remove features having information values that satisfy a predetermined threshold condition.

4. The method for building a data identification model of claim 1 , wherein prior to selecting the second training sample set from positive samples that have identification results after being identified using the first model, the method further comprises:

performing DNN training using the first training sample set to obtain a second model.

5. The method for building a data identification model of claim 4 , wherein selecting the second training sample set from positive samples that have identification results after being identified using the first model further comprises:

evaluating the first model and obtaining an ROC curve corresponding to the first model;

evaluating the second model and obtaining an ROC curve corresponding to the second model; and

selecting, based on a threshold probability corresponding to an intersection point of the ROC curves of the first model and the second model, samples having probabilities satisfying a threshold probability condition from positive samples that have identification results after being identified using the first model to serve as the second training sample set.

6. An apparatus for building a data identification model, comprising:

a memory storing a set of instructions; and

a processor configured to execute the set of instructions to cause the apparatus to perform:

logistic regression training using training samples to obtain a first model, the training samples comprising positive samples and negative samples;

sampling the training samples proportionally to obtain a first training sample set;

identifying the positive samples using the first model, and selecting a second training sample set from positive samples that have identification results after being identified using the first model; and

Deep Neural Networks (DNN) training using the first training sample set and the second training sample set to obtain a final data identification model.

7. The apparatus for building a data identification model of claim 6 , wherein the processor is configured to execute the set of instructions to cause the apparatus to further perform:

prior to sampling the training samples proportionally or performing logistic regression training, feature engineering preprocessing on the training samples.

8. The apparatus for building a data identification model of claim 7 , wherein the processor is configured to execute the set of instructions to cause the apparatus to further perform:

prior to performing logistic regression training using the training samples to obtain a first model, feature screening on the training samples to remove features having information values that satisfy a predetermined threshold condition.

9. The apparatus for building a data identification model of claim 6 , wherein the processor is configured to execute the set of instructions to cause the apparatus to further perform:

DNN training using the first training sample set to obtain a second model.

10. The apparatus for building a data identification model of claim 9 , wherein selecting the second training sample set from positive samples that have identification results after being identified using the first model further comprises:

evaluating the first model and obtaining an ROC curve corresponding to the first model;

evaluating the second model and obtaining an ROC curve corresponding to the second model; and

selecting, based on a threshold probability corresponding to an intersection point of the ROC curves of the first model and the second model, samples having probabilities satisfying a threshold probability condition from positive samples that have identification results after being identified using the first model to serve as the second training sample set.

11. A non-transitory computer readable medium that stores a set of instructions that is executable by at least one processor of a computer to cause the computer to perform a method for building a data identification model, the method comprising:

performing logistic regression training using training samples to obtain a first model, the training samples comprising positive samples and negative samples;

sampling the training samples proportionally to obtain a first training sample set;

identifying the positive samples using the first model, and selecting a second training sample set from positive samples that have identification results after being identified using the first model; and

performing Deep Neural Networks (DNN) training using the first training sample set and the second training sample set to obtain a final data identification model.

12. The non-transitory computer readable medium of claim 11 , wherein the processor is configured to execute the set of instructions to cause the computer to further perform:

prior to sampling the training samples proportionally or performing logistic regression training, feature engineering preprocessing on the training samples.

13. The non-transitory computer readable medium of claim 12 , wherein the processor is configured to execute the set of instructions to cause the computer to further perform:

prior to performing logistic regression training using the training samples to obtain a first model, feature screening on the training samples to remove features having information values that satisfy a predetermined threshold condition.

14. The non-transitory computer readable medium of claim 13 , wherein the predetermined threshold condition is satisfied when the features have information values less than a predetermined threshold.

15. The non-transitory computer readable medium of claim 11 , wherein the processor is configured to execute the set of instructions to cause the computer to further perform:

DNN training using the first training sample set to obtain a second model.

16. The non-transitory computer readable medium of claim 15 , selecting the second training sample set from positive samples that have identification results after being identified using the first model further comprises:

evaluating the first model and obtain an ROC curve corresponding to the first model;

evaluating the second model and obtain an ROC curve corresponding to the second model; and

selecting, based on a threshold probability corresponding to an intersection point of the ROC curves of the first model and the second model, samples having probabilities satisfying a threshold probability condition from positive samples that have identification results after being identified using the first model to serve as the second training sample set.

17. The non-transitory computer readable medium of claim 16 , wherein the threshold probability condition is satisfied when the samples have probabilities less than a threshold probability.