Optimized interchange code prediction system for processing credit card transactions
A computer-implemented method for predicting interchange charges includes: retrieving a historical transactions set, where each completed transaction in the set includes transaction features, a bank identification number (BIN), and a corresponding true interchange code; transforming all BINs in the set into a corresponding plurality of BIN features that comprise probabilities; creating a first training set including all transaction features, all pluralities of BIN features, and all true interchange codes associated with the historical transactions set; training a random forest model using the first training set and generating a second training set including rounded BIN features, rounded transaction features, discrete ones of the transaction features, and the true interchange codes; training the random forest model using the second training set to generate a trained random forest model for prediction of the interchange codes; and executing the trained random forest model for new transactions to generate corresponding predicted interchange codes.
1. A computer implemented method for predicting interchange codes corresponding to credit card transactions, the method comprising:
via a server:
receiving transaction data corresponding to a transaction from a merchant point-of-sale terminal;
extracting transaction-specific features from the transaction data;
translating and rounding non-discrete values of the transaction-specific features into rounded transaction-specific features according to cut points determined by prior execution of a trained neural network to predict interchange charges for historical transactions associated with a plurality of merchants using historical transaction features as inputs, and wherein outputs of the trained neural network comprise predicted interchange charges for all combinations of the inputs, and wherein a lookup table that maps each of the all combinations of the inputs to a corresponding predicted interchange charge is generated during the prior execution of the trained neural network;
using the rounded transaction-specific features to index the lookup table to obtain a transaction-specific predicted interchange charge for the transaction wherein:
the transaction-specific predicted interchange charge comprises a weighted sum of predicted interchange charges for a subset of all of the interchange charges;
a weight for each term of the weighted sum indicates a probability that an associated one of the combinations of the set of all rounded transaction features will result in a corresponding predicted interchange charge within the subset; and
probabilities for weights in the weighted sum are previously determined by executing the trained neural network using historical rounded transaction features;
computing a holdback amount for the transaction based upon the transaction-specific predicted interchange charge; and
transmitting the transaction data and holdback amount to a payment processor for settlement of the transaction.
2. The computer-implemented method as recited in claim 1 , wherein the transaction features further comprise a total transaction amount along and a tip percentage.
3. The computer-implemented method as recited in claim 2 , wherein the transaction features further comprise a type of transaction.
4. The computer-implemented method as recited in claim 3 , wherein the transaction features father comprise an address verification system response.
5. The computer-implemented method as recited in claim 4 , wherein the transaction features father comprise a card network, a card type, and a merchant category code.
6. The computer-implemented method as recited in claim 1 , wherein the transaction features comprise a fixed number of bank identification number (BIN) features that are determined by statistically analyzing all BINS to determine probabilities that a given BIN exhibits a corresponding one of the BIN features.
7. The computer-implemented method as recited in claim 1 , wherein the fixed number of BIN features comprise regulated debit, debit, rewards, and high rewards.
8. A computer-readable storage median storing instructions that, when executed by a computer, came the computer to perform a method for using a computer system to predict interchange codes corresponding to credit card transactions, the method comprising:
via a server:
receiving transaction data corresponding to a transaction from a merchant point-of-sale terminal;
extracting transaction-specific features from the transaction data;
translating and rounding non-discrete values of the transaction-specific features into rounded transaction-specific features according to cut points determined by prior execution of a trained neural network to predict interchange charges for historical transactions associated with a plurality of merchants using historical transaction features as inputs, and wherein outputs of the trained neural network comprise predicted interchange charges for all combinations of the inputs, and wherein a lookup table that maps each of the all combinations of the inputs to a corresponding predicted interchange charge is generated during the prior execution of the trained neural network;
using the rounded transaction-specific features to index the lookup table to obtain a transaction-specific predicted interchange charge for the transaction, wherein:
the transaction-specific predicted interchange charge comprises a weighted sum of predicted interchange charges fora subset of all of the interchange charges;
a weight for each term of the weighted sum indicates a probability that an associated one of the combinations of the set of all rounded transaction features will result in a corresponding predicted interchange charge within the subset; and
probabilities for weights in the weighted sun are previously determined by executing the trained neural network using historical rounded transaction features;
computing a holdback amount for the transaction based upon the transaction-specific predicted interchange charge; and
transmitting the transaction data and holdback amount to a payment processor for settlement of the transaction.
9. The computer-readable storage medium as recited in claim 8 , wherein the transaction features tither comprise a total transaction amount along and a tip percentage.
10. The computer-readable storage medium as recited in claim 9 , wherein the transaction features further comprise a type of transaction.
11. The computer-readable storage medium as recited in claim 10 , wherein the transaction features father comprise an address verification system response.
12. The computer-readable storage medium as recited in claim 11 , wherein the transaction features further comp rise a card network, a card type, and a merchant category code.
13. The computer-readable storage medium as recited in claim 8 , wherein the transaction features comprise a fixed number of bank identification number (BIN) features that are determined by statistically analyzing all BINS to determine probabilities that a given BIN exhibits a corresponding one of the BIN features.
14. The computer-readable storage medium as recited in claim 8 , wherein the fixed number of BIN features comprise regulated debit, debit, rewards, and high rewards.
15. A system for predicting interchange codes corresponding to credit card transactions, the system comprising:
a server, configured to:
receive transaction data corresponding to a transaction from a merchant point-of-sale terminal;
extract transaction-specific features from the transaction data;
translate and round non-discrete values of the transaction-specific features into rounded transaction-specific features according to cut points determined by prior execution of a trained neural network to predict interchange charges for historical transactions associated with a plurality of merchants using historical transaction features as inputs, and wherein outputs of the trained neural network comprise predicted interchange charges for all combinations of the inputs, and wherein a lookup table that maps each of the all combinations of the inputs to a corresponding predicted interchange charge is generated during the prior execution of the trained neural network;
using the rounded transaction-specific features to index the lookup table to obtain a transaction-specific predicted interchange charge for the transaction, wherein:
the transaction-specific predicted interchange charge comprises a weighted sum of predicted interchange charges for a subset of all of the interchange charges;
a weight for each term of the weighted sum indicates a probability that an associated one of the combinations of the set of all rounded transaction features will result in a corresponding predicted interchange charge within the subset; and
probabilities for weights in the weighted sun are previously determined by executing the trained natal network using historical rounded transaction features;
computing a holdback amount for the transaction based upon the transaction-specific predicted interchange charge; and
transmitting the transaction data and holdback amount to a payment processor for settlement of the transaction.
16. The system as recited in claim 15 , wherein the transaction features further comprise a total transaction amount along and a tip percentage.
17. The system as recited in claim 16 , wherein the transaction features further comprise a type of transaction.
18. The system as recited in claim 17 , wherein the transaction features further comprise an address verification system response.
19. The system as recited in claim 15 , wherein the transaction features comprise a fixed number of bank identification number (BIN) features that are determined by statistically analyzing all BINS to determine probabilities that a given BIN exhibits a corresponding one of the BIN features.
20. System as recited in claim 15 , wherein the fixed number of BIN features comprise regulated debit, debit, rewards, and high rewards.