Insufficient funds predictor

Patent metadata

Field	Value
Publication number	US-11042930-B1
Application number	US-201815925817-A
Country	US
Kind code	B1
Filing date	Mar 20, 2018
Priority date	Mar 20, 2018
Publication date	Jun 22, 2021
Grant date	Jun 22, 2021

Abstract

Certain aspects of the present disclosure provide techniques for improving a prediction of whether a non-sufficient funds fee will be incurred by a user utilizing machine learning techniques. For example, a predictive model may be trained using machine learning techniques based on historical data and derived data for a plurality of users. The predictive model may then be used to predict a probability of a particular user incurring an insufficient funds fee. The probability of the particular user may be used to generate an alert and suggestion to be presented to the particular user to avoid incurring the insufficient funds fee.

First claim

What is claimed is:

1. A system comprising:
a processor; and
a memory comprising executable instructions, which, when executed by the processor, cause the system to:
train a plurality of predictive models using training data from a plurality of backward-looking periods to generate predictions for a given forward-looking period;
determine accuracies of the plurality of predictive models using testing data by comparing outputs from the plurality of predictive models to labels in the testing data indicating whether insufficient fund fees were historically incurred;
select a backward-looking period of the plurality of backward-looking periods that is most accurate for the given forward-looking period based on which accuracy of the accuracies is highest;
obtain historical data corresponding to the selected backward looking period and related to a financial history of each of a plurality of users;
determine one or more historical data variables from the historical data;
generate one or more derived variables from the historical data;
form a plurality of data subsets from the historical data based on the one or more historical data variables and the one or more derived variables, each data subset of the plurality of data subsets comprising a different set of data than any other data subset in the plurality of data subsets;
train a boosted decision tree model based on the plurality of data subsets;
obtain new financial data related to a user;
predict a probability of the user incurring an insufficient funds fee during the given forward-looking period with the boosted decision tree model;
determine, based on the probability, a risk level that the user will incur the insufficient funds fee during the given forward-looking period, wherein the risk level is selected from a plurality of risk levels comprising:
a low risk level that is associated with a first probability range;
a medium risk level that is associated with a second probability range; and
a high risk level that is associated with a third probability range; and

determine a notification to provide to the user based on the risk level.

2. The system of claim 1 , the memory further comprising executable instructions, which, when executed by the processor, cause the system to present the notification to the user to alert the user of the probability of incurring the insufficient funds fee.

3. The system of claim 2 , wherein the notification comprises a suggested action for the user to avoid the insufficient funds fee.

4. The system of claim 1 , the memory further comprising executable instructions, which, when executed by the processor, cause the system to:
generate a report of the risk level that the user will incur the insufficient funds fee, the report including a historical visualization of the risk level of the user; and
present the report to the user.

5. The system of claim 1 , the memory further comprising executable instructions, which, when executed by the processor, cause the system to:
generate a plurality of decision tree-based models based on the plurality of data subsets using a gradient boosting technique; and
generate the boosted decision tree model based on the plurality of decision tree-based models.

6. The system of claim 1 , wherein the plurality of users comprises only users that have incurred an insufficient funds fee during the backward-looking period.

7. A non-transitory computer-readable storage medium storing instructions, which, when executed on a processor, performs an operation for an insufficient funds fee, the operation comprising:
training a plurality of predictive models using training data from a plurality of backward-looking periods to generate predictions for a given forward-looking period;
determining accuracies of the plurality of predictive models using testing data by comparing outputs from the plurality of predictive models to labels in the testing data indicating whether insufficient fund fees were historically incurred;
selecting a backward-looking period of the plurality of backward-looking periods that is most accurate for the given forward-looking period based on which accuracy of the accuracies is highest;
obtaining historical data corresponding to the selected backward looking period and related to a financial history of each of a plurality of users;
determining one or more historical data variables from the historical data;
generating one or more derived variables from the historical data;
forming a plurality of data subsets from the historical data based on the one or more historical data variables and the one or more derived variables, each data subset of the plurality of data subsets comprising a different set of data than any other data subset in the plurality of data subsets;
training a boosted decision tree model based on the plurality of data subsets;
obtaining new financial data related to a user;
predicting a probability of the user incurring an insufficient funds fee during the given forward-looking period with the boosted decision tree model;
determining, based on the probability, a risk level that the user will incur the insufficient funds fee during the given forward-looking period, wherein the risk level is selected from a plurality of risk levels comprising:
a low risk level that is associated with a first probability range;
a medium risk level that is associated with a second probability range; and
a high risk level that is associated with a third probability range; and

determining a notification to provide to the user based on the risk level.

8. The non-transitory computer-readable storage medium of claim 7 , the operation further comprising: presenting the notification to the user to alert the user of the probability of incurring the insufficient funds fee.

9. The non-transitory computer-readable storage medium of claim 8 , wherein the notification comprises a suggested action for the user to avoid the insufficient funds fee.

10. The non-transitory computer-readable storage medium of claim 7 , the operation further comprising:
generating a report of the risk level that the user will incur the insufficient funds fee, the report including a historical visualization of the risk level of the user; and
presenting the report to the user.

11. The non-transitory computer-readable storage medium of claim 7 , the operation further comprising:
generating a plurality of decision tree-based models based on the plurality of data subsets using a gradient boosting technique; and
generating the boosted decision tree model based on the plurality of decision tree-based models.

12. The non-transitory computer-readable storage medium of claim 7 , wherein the plurality of users comprises only users that have incurred an insufficient funds fee during the backward-looking period.

13. The non-transitory computer-readable storage medium of claim 7 , wherein each data subset in the plurality of data subsets comprises respective historical data related to one or more respective users of the plurality of users.

14. A method for predicting an insufficient funds fee, the method comprising:
training a plurality of predictive models using training data from a plurality of backward-looking periods to generate predictions for a given forward-looking period;
determining accuracies of the plurality of predictive models using testing data by comparing outputs from the plurality of predictive models to labels in the testing data indicating whether insufficient fund fees were historically incurred;
selecting a backward-looking period of the plurality of backward-looking periods that is most accurate for the

Assignees

• Intuit Inc

Inventors

• Mintz Ido
• Hayman Liron
• Mishraky Elhanan

Classifications

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N20/20

  Ensemble learning · CPC title

• G06Q40/02 Primary

  Banking, e.g. interest calculation or account maintenance (credit or loans G06Q40/03) · CPC title

• G06N7/005

  Physics · mapped topic