Back to results
US 12,614,180 B1Granted
Dynamic prediction of transaction properties
AssigneeCapital One Services, LLC
InventorsJoshua Edwards, Viral Jogani, Oliver Deighton, Yao Zhang, Elizabeth Garner
Life of this patent
Estimating the term…
Abstract
Systems, methods, and apparatuses are described for predicting the identity of significant transactions. A plurality of different transactions may be associated with an account. Those transactions may be filtered and processed (e.g., using a trained machine learning model) to identify one or more significant transactions, such as recurring transactions. Such recurring transaction identification may be performed to generate rules for identifying future recurring transactions. Additionally and/or alternatively, a second machine learning model may be trained to predict the value of recurring transactions in a region.
Claims20
A computing device configured to automatically identify high-value recurring transactions based on historical transaction data, the computing device comprising:
- one or more processors;
- and
- memory storing instructions that, when executed by the one or more processors, cause the computing device to:
- determine a first plurality of transactions corresponding to an account associated with a geographical region;
- filter the first plurality of transactions by:
- removing, from the first plurality of transactions, one or more first transactions based on a Merchant Category Code (MCC) corresponding to the one or more first transactions;
- and
- removing, from the first plurality of transactions, one or more second transactions based on an amount of the one or more second transactions;
- provide, as input to a first trained machine learning model, at least a portion of the filtered first plurality of transactions, wherein the first trained machine learning model comprises a first artificial neural network trained, using first training data comprising a history of known recurring transactions corresponding to the account, to identify correlations in transactions;
- receive, as output from the first trained machine learning model, an identification of at least two transactions of the filtered first plurality of transactions;
- provide, as input to a second trained machine learning model, the at least two transactions of the filtered first plurality of transactions, wherein the second trained machine learning model comprises a second artificial neural network trained by:
- performing Optical Character Recognition (OCR) on a plurality of document images to identify a plurality of descriptive text elements;
- identifying a plurality of distance values by calculating, using a natural language processing algorithm, a distance between one or more words of each of the plurality of descriptive text elements and one or more target words;
- generating second training data that indicates associations between the descriptive text elements and the plurality of distance values;
- and
- modifying, based on the second training data, one or more weights of one or more nodes of the second artificial neural network;
- receive, as output from the second trained machine learning model, an identification of a transaction from the at least two transactions of the filtered first plurality of transactions;
- generate, based on one or more properties of the transaction, one or more rules configured to predict high-volume recurring transaction amounts in the geographical region;
- and
- identify, based on the one or more rules, a first high-volume recurring transaction amount for a first geographic region.
17 more elements
A method configured to automatically identify high-value recurring transactions based on historical transaction data, the method comprising:
- determining a first plurality of transactions corresponding to an account associated with a geographical region;
- filtering the first plurality of transactions by:
- removing, from the first plurality of transactions, one or more first transactions based on a Merchant Category Code (MCC) corresponding to the one or more first transactions;
- and
- removing, from the first plurality of transactions, one or more second transactions based on an amount of the one or more second transactions;
- provide, as input to a first trained machine learning model, at least a portion of the filtered first plurality of transactions, wherein the first trained machine learning model comprises a first artificial neural network trained, using first training data comprising a history of known recurring transactions corresponding to the account, to identify correlations in transactions;
- receive, as output from the first trained machine learning model, an identification of at least two transactions of the filtered first plurality of transactions;
- providing, as input to a second trained machine learning model, the at least two transactions of the filtered first plurality of transactions, wherein the second trained machine learning model comprises a second artificial neural network trained by:
- performing Optical Character Recognition (OCR) on a plurality of document images to identify a plurality of descriptive text elements;
- identifying a plurality of distance values by calculating, using a natural language processing algorithm, a distance between one or more words of each of the plurality of descriptive text elements and one or more target words;
- generating second training data that indicates associations between the descriptive text elements and the plurality of distance values;
- and
- modifying, based on the second training data, one or more weights of one or more nodes of the second artificial neural network;
- receiving, as output from the second trained machine learning model, an identification of a transaction from the at least two transactions of the filtered first plurality of transactions;
- generating, based on one or more properties of the transaction, one or more rules configured to predict high-volume recurring transaction amounts in the geographical region;
- and
- identifying, based on the one or more rules, a first high-volume recurring transaction amount for a first geographic region.
14 more elements
One or more non-transitory computer-readable media storing instructions configured to automatically identify high-value recurring transactions based on historical transaction data, wherein the instructions, when executed by one or more processors of a computing device, cause the computing device to:
- determine a first plurality of transactions corresponding to an account associated with a geographical region;
- filter the first plurality of transactions by:
- removing, from the first plurality of transactions, one or more first transactions based on a Merchant Category Code (MCC) corresponding to the one or more first transactions;
- and
- removing, from the first plurality of transactions, one or more second transactions based on an amount of the one or more second transactions;
- provide, as input to a first trained machine learning model, at least a portion of the filtered first plurality of transactions, wherein the first trained machine learning model comprises a first artificial neural network trained, using first training data comprising a history of known recurring transactions corresponding to the account, to identify correlations in transactions;
- receive, as output from the first trained machine learning model, an identification of at least two transactions of the filtered first plurality of transactions;
- provide, as input to a second trained machine learning model, the at least two transactions of the filtered first plurality of transactions, wherein the second trained machine learning model comprises a second artificial neural network trained by:
- performing Optical Character Recognition (OCR) on a plurality of document images to identify a plurality of descriptive text elements;
- identifying a plurality of distance values by calculating, using a natural language processing algorithm, a distance between one or more words of each of the plurality of descriptive text elements and one or more target words;
- generating second training data that indicates associations between the descriptive text elements and the plurality of distance values;
- and
- modifying, based on the second training data, one or more weights of one or more nodes of the second artificial neural network;
- receive, as output from the second trained machine learning model, an identification of;
- generate, based on one or more properties of the transaction, one or more rules configured to predict high-volume recurring transaction amounts in the geographical region;
- and
- identify, based on the one or more rules, a first high-volume recurring transaction amount for a first geographic region.
14 more elements
Dependencies are inferred from the claim text.