Adaptive machine learning threshold
In an example embodiment, a solution is provided wherein a threshold used by a classifier in a first machine learning model is dynamically set by a second machine learning model. More particularly, the threshold may be selected from two or more different threshold settings, based on the output of the second machine learning model. This acts to improve the reliability of predictions made by the first machine learning model in certain use cases where circumstances not adequately captured by the first machine learning model can affect the accuracy of the threshold used by the first machine learning model.
1. A method for using a first neural network to improve operation of a second neural network, comprising:
training, by one or more processors, the first neural network based on a set of transaction data for a training transaction labeled with an indication that an enumeration attack occurred on a merchant computing device at the time that a validation of the training transaction was attempted;
training, by the one or more processors, the second neural network based on the set of transaction data used to train the first neural network, wherein the set of transaction data are labeled with an indication that the training transaction was validated by an issuer computing device;
receiving, by the one or more processors, transaction data for a credit card transaction initiated at the merchant computing device;
determining, by the one or more processors using the first neural network, the merchant computing device is under a first enumeration attack using the transaction data for the credit card transaction initiated at the merchant computing device;
adjusting, by the one or more processors, a confidence score threshold of the second neural network based on the determination by the first neural network that the merchant computing device is under the first enumeration attack;
determining, by the one or more processors using a second neural network, a confidence score that the credit card transaction will be validated by the issuer computing device; and
blocking, by the one or more processors, the credit card transaction responsive to determining the confidence score is below the adjusted confidence score threshold of the second neural network.
2. The method of claim 1 , further comprising training the first neural network and training of the second neural network using the same training data but with different labels.
3. The method of claim 2 , wherein the training data used for the training of the second neural network has labels obtained from actual approvals and declines of corresponding validations in the training data by the issuer computing device.
4. The method of claim 2 , wherein the training data includes features selected from the group comprising: time features, customer data, client data, and card metadata.
5. A system comprising:
one or more processors; and
at least one memory storing instructions that, when executed by at least one processor among the one or more processors, cause the at least one processor to perform operations comprising:
training a first neural network based on a set of transaction data for a training transaction labeled with an indication that an enumeration attack occurred on a merchant computing device at the time that a validation of the training transaction was attempted;
training a second neural network based on the set of transaction data used to train the first neural network, wherein the set of transaction data are labeled with an indication that the training transaction was validated by an issuer computing device;
receiving transaction data for a credit card transaction initiated at the merchant computing device;
determining, using the first neural network, the merchant computing device is under a first enumeration attack using the transaction data for the credit card transaction initiated at the merchant computing device;
adjusting a threshold based on the determination that the merchant computing device is under the first enumeration attack;
determining, using a second neural network, a confidence score that the credit card transaction will be validated by the issuer computing device; and
blocking the credit card transaction responsive to determining the confidence score is below the adjusted confidence score threshold.
6. The system of claim 5 , further comprising training the first neural network and training the second neural network using the same training data but with different labels.
7. The system of claim 6 , wherein the training data used for the training of the second neural network has labels obtained from actual approvals and declines of corresponding validations in the training data by the issuer computing device.
8. The system of claim 6 , wherein the training data includes features selected from the group comprising: time features, customer data, client data, and card metadata.
9. A non-transitory machine-readable medium comprising instructions which, when read by a machine, cause the machine to perform operations comprising:
training a first neural network based on a set of transaction data for a training transaction labeled with an indication that an enumeration attack occurred on a merchant computing device at the time that a validation of the training transaction was attempted;
training a second neural network based on the set of transaction data used to train the first neural network, wherein the set of transaction data are labeled with an indication that the training transaction was validated by an issuer computing device;
receiving transaction data for a credit card transaction initiated at the merchant computing device;
determining, using the first neural network, the merchant computing device is under a first enumeration attack using the transaction data for the credit card transaction initiated at the merchant computing device;
adjusting a confidence score threshold based on the determination that the merchant computing device is under the first enumeration attack;
determining, using the second neural network, a confidence score that the credit card transaction will be validated by the issuer computing device; and
blocking the credit card transaction responsive to determining the confidence score is below the adjusted confidence score threshold.
10. The non-transitory machine-readable medium of claim 9 , further comprising training the first neural network and training the second neural network using the same training data but with different labels.
11. The non-transitory machine-readable medium of claim 10 , wherein the training data used for the training of the second neural network has labels obtained from actual approvals and declines of corresponding validations in the training data by the issuer computing device.
12. The non-transitory machine-readable medium of claim 10 , wherein the training data includes features selected from the group comprising: time features, customer data, client data, and card metadata.