Method, apparatus and electronic device for identifying risks pertaining to transactions to be processed

1. A method for identifying a risk pertaining to a transaction to be processed on an online platform, comprising:

generating, in a fast identification layer in a risk control system, a model for identifying transactions susceptible to false-positive interference, wherein the model is generated by machine learning based on positive samples and negative samples, the machine learning including at least one of decision tree-based machine learning or in-depth learning based on neural networks, the positive samples including interfered transactions for training the model for identifying transactions susceptible to false-positive interference, the negative samples including non-interfered transactions that are subject to no risks for training the model for identifying transactions susceptible to false-positive interference, wherein when the model for identifying transactions susceptible to false-positive interference is in an initial state, the positive samples are obtained by weighted sampling all transactions in a specified historical period, wherein a weight of weighted sampling is negatively correlated to an interval between a sampling time and a time of a sampled transaction;

identifying a risk pertaining to the transaction on the online platform via the fast identification layer in the risk control system, thereby obtaining a first identification result;

identifying a risk of false-positive interference pertaining to the transaction on the online platform via the model for identifying transactions susceptible to false-positive interference, thereby obtaining a second identification result; and

determining a processing manner for the transaction on the online platform based on the first identification result and the second identification result, the processing manner including releasing the transaction on the online platform, or further identifying the risk pertaining to the transaction on the online platform via an in-depth identification layer in the risk control system,

wherein the risk control system is implemented by a server, and the generating the model for identifying transactions susceptible to false-positive interference, the identifying the risk pertaining to the transaction on the online platform, the identifying the risk of false-positive interference pertaining to the transaction on the online platform, and the determining the processing manner for the transaction on the online platform are performed by a processor of the server.

2. The method according to claim 1 , wherein weights of the positive samples comply with formula w i =e −na , where w i , is the weight of an i th positive sample, η is a predefined adjustable parameter, and α i , is an interval of the i th positive sample.

3. The method according to claim 1 , wherein the negative samples are obtained by evenly sampling in all transactions in the specified historical period.

4. The method according to claim 1 , wherein when the model for identifying transactions susceptible to false-positive interference is a newer version model obtained by updating an older version model, the positive samples comprise at least one of: interfered transactions not identified in the older version model, or interfered transactions identified in the older version model.

5. The method according to claim 4 , wherein the negative samples are obtained by evenly sampling in all transactions in a period in which the positive samples are obtained.

6. The method according to claim 1 , wherein a version of the model for identifying transactions susceptible to false-positive interference is updated according to a preset period.

7. The method according to claim 1 , wherein determining the processing manner of the transaction on the online platform based on the first identification result and the second identification result comprises:

calculating a score of a decision function according to f(R,D)=R·I(β>R>α)·e −D +I(R≥β), where f(R,D) is the score of the decision function, R is a risk score included in the first identification result, D is an interference tendency score included in the second identification result, α and β are preset values and 0<(α,β)<1, e is a base of a natural exponential function, and a value of I(x) is 1 when condition x is met and 0 when condition x is not met;

when f(R,D)<θ, the processing manner is releasing the transaction on the online platform; and

when θ<f(R,D)≤1, the in-depth identification layer in the risk control system further identifies the risk pertaining to the transaction on the online platform;

wherein θ is a preset threshold value, and 0<θ<1.

8. An electronic device operating as a server, comprising:

a processor; and

a memory storing instructions executable by the processor;

wherein the processor is configured to:

generate, in a fast identification layer in a risk control system, a model for identifying transactions susceptible to false-positive interference, wherein the model is generated by machine learning based on positive samples and negative samples, the machine learning including at least one of decision tree-based machine learning or in-depth learning based on neural networks, the positive samples including interfered transactions for training the model for identifying transactions susceptible to false-positive interference, the negative samples including non-interfered transactions that are subject to no risks for training the model for identifying transactions susceptible to false-positive interference, wherein when the model for identifying transactions susceptible to false-positive interference is in an initial state, the positive samples are obtained by weighted sampling all transactions in a specified historical period, wherein a weight of weighted sampling is negatively correlated to an interval between a sampling time and a time of a sampled transaction;

identify a risk pertaining to a transaction to be processed on an online platform via the fast identification layer in the risk control system, thereby obtaining a first identification result;

identify a risk of false-positive interference pertaining to the transaction on the online platform via the model for identifying transactions susceptible to false-positive interference, thereby obtaining a second identification result; and

determine a processing manner for the transaction on the online platform based on the first identification result and the second identification result, the processing manner including releasing the transaction on the online platform, or further identifying the risk pertaining to the transaction on the online platform via an in-depth identification layer in the risk control system,

wherein the risk control system is implemented by the server.

9. The electronic device according to claim 8 , wherein weights of the positive samples comply with formula w i =e −nα , where w i , is the weight of an i th positive sample, η is a predefined adjustable parameter, and α i , is an interval of the i th positive sample.

10. The electronic device according to claim 8 , wherein the negative samples are obtained by evenly sampling in all transactions within a specified historical period.

11. The electronic device according to claim 8 , wherein when the model for identifying transactions susceptible to false-positive interference is a newer version model obtained by updating an older version model, the positive samples include at least one of: interfered transactions not identified in the older version model, interfered transactions identified in the older version model.

12. The electronic device according to claim 11 , wherein the negative samples are obtained by evenly sampling in all transactions within a period in which the positive samples are obtained.

13. The electronic device according to claim 8 , wherein a version of the model for identifying transactions susceptible to false-positive interference is updated according to a preset period.

14. The electronic device according to claim 8 , wherein the processor is further configured to: calculate a score of a decision function according to f(R,D)=R·I(β>R>α)·e −D +I(R≥β), where f(R,D) is the score of the decision function, R is a risk score included in the first identification result, D is an interference tendency score included in the second identification result, α and β are preset values and 0<(α,β)<1, e is a base of a natural exponential function, and the value of I(x) is 1 when condition x is met and 0 when condition x is not met;

when f(R,D)<θ, the processing manner is releasing the transaction on the online platform; and

when θ<(f(R,D)≤1, the in-depth identification layer in the risk control system further identifies the risk pertaining to the transaction on the online platform,

wherein θ is a preset threshold value, and 0<θ<1.

15. A non-transitory computer-readable storage medium having stored therein instructions that, when executed by a processor of a device operating as a server, cause the device to perform a method for identifying a risk pertaining to a transaction to be processed on an online platform, the method comprising:

generating, in a fast identification layer in a risk control system, a model for identifying transactions susceptible to false-positive interference, wherein the model is generated by machine learning based on positive samples and negative samples, the machine learning including at least one of decision tree-based machine learning or in-depth learning based on neural networks, the positive samples including interfered transactions for training the model for identifying transactions susceptible to false-positive interference, the negative samples including non-interfered transactions that are subject to no risks for training the model for identifying transactions susceptible to false-positive interference, wherein when the model for identifying transactions susceptible to false-positive interference is in an initial state, the positive samples are obtained by weighted sampling all transactions in a specified historical period, wherein a weight of weighted sampling is negatively correlated to an interval between a sampling time and a time of a sampled transaction;

identifying a risk pertaining to the transaction on the online platform via the fast identification layer in the risk control system, thereby obtaining a first identification result;

identifying a risk of false-positive interference pertaining to the transaction on the online platform via the model for identifying transactions susceptible to false-positive interference, thereby obtaining a second identification result; and

determining a processing manner for the transaction on the online platform based on the first identification result and the second identification result, the processing manner including releasing the transaction on the online platform, or further identifying the risk pertaining to the transaction on the online platform via an in-depth identification layer in the risk control system, wherein the risk control system is implemented by the server.