Methods and systems for efficiently retrieving financial data for financial accounts via a comunication network

Patent metadata

Field	Value
Publication number	US-10861092-B1
Application number	US-201715795110-A
Country	US
Kind code	B1
Filing date	Oct 26, 2017
Priority date	Oct 26, 2017
Publication date	Dec 8, 2020
Grant date	Dec 8, 2020

Abstract

Computerized systems and methods for efficiently refreshing financial data for financial accounts for login identifications of users at respective financial institutions in a batch data pull via a communication network. The login identifications are prioritized for being refreshed based on determining a probability that each login identification has a new transaction since the last successful refresh of the login identification using a learning algorithm utilizing past financial data. The login identifications are then refreshed in a batch data pull from the financial institutions in an order from highest probability to lowest probability of having a new transaction since the last batch data pull.

First claim

What is claimed is:

1. A computer-implemented method for retrieving financial data for financial accounts of users via a communication network using each user's online login identifications for the financial accounts, comprising:
storing a plurality of online login identifications for financial accounts for a plurality of users in a database;
determining a last successful data retrieval time for each online login identification in which financial data was successfully retrieved for the respective online login identification;
training a plurality of predictive models, wherein each predictive model is associated with a different model selection time period, said training comprising:
a. assigning each online login identification to a respective predictive model having a last successful data retrieval time corresponding to the model selection time period associated with the predictive model;
b. for each of the plurality of predictive models,
i. accessing financial data for each of the online login identifications assigned to the respective predictive model for a plurality of data time periods prior to the last successful data retrieval time;
ii. generating a respective model training set of data by determining whether there was a new transaction in each of the plurality of data time periods for each respective online login identification assigned to the respective predictive model using the financial data;
iii. training the respective predictive model by inputting the respective model training set into a respective learning algorithm for the respective predictive model;

for each online login identification, determining a probability score that such online login identification has a new transaction since the last successful data retrieval time by inputting the respective determination of whether there was a new transaction in each of the plurality of data time periods for each respective online login identification into the trained predictive model assigned to the online login identification;
retrieving data for a plurality of the online login identifications via an online connection in an order of the probability scores from higher probability of having a new transaction to lower probability of having a new transaction since the last successful data retrieval time.

2. The method of claim 1 , further comprising:
sorting the online login identifications based on the probability scores from higher probability of having a new transaction to lower probability of having a new transaction since the last successful data retrieval time; and
wherein retrieving data for a plurality of the online login identifications is based on the sorted online login identifications.

3. The method of claim 1 , wherein the data time periods are days and the plurality of data time periods is a plurality of consecutive days.

4. The method of claim 3 , wherein the plurality of consecutive days is at least 30 days.

5. The method of claim 3 , wherein the plurality of consecutive days is at least 90 day.

6. The method of claim 1 , wherein the learning algorithms are each random forest machine learning algorithms.

7. The method of claim 1 , wherein retrieving data for a plurality of the online login identifications is performed for a predetermined percentage of the online login identifications.

8. The method of claim 7 , wherein the predetermined percentage is selected from the group consisting of 90%, 80%, 70%, 60%, 50%, 40%, 30%, 25%, 20% or 15%.

9. The method of claim 7 , wherein the predetermined percentage is selected to result in a percentage of login identifications which are current after completing the step of retrieving data for a plurality of the online login identifications of at least 70%.

10. A system for retrieving financial data for a plurality of financial accounts, comprising
a computer server having a computer processor, a data store storing financial data for a plurality of financial accounts for a plurality of users and a plurality of online login identifications for the financial accounts for the users, a network communication interface, and a financial data retrieval software application, the computer server in communication with a plurality of financial institutions via a communication network;
the financial data retrieval software application configured to program the computer to perform the following process:
determining a last successful data retrieval time for each online login identification in which financial data was successfully retrieved for the respective online login identification;
training a plurality of predictive models, wherein each predictive model is associated with a different model selection time period said training comprising:
a. assigning each online login identification to a respective predictive model having a last successful data retrieval time corresponding to the model selection time period associated with the predictive model;
b. for each of the plurality of predictive models,
i. accessing financial data for each of the online login identifications assigned to the respective predictive model for a plurality of data time periods prior to the last successful data retrieval time;
ii. generating a respective model training set of data by determining whether there was a new transaction in each of the plurality of data time periods for each respective online login identification assigned to the respective predictive model using the financial data;
iii. training the respective predictive model by inputting the respective model training set into a respective learning algorithm for the respective predictive model;

for each online login identification, determining a probability score that such online login identification has a new transaction since the last successful data retrieval time by inputting the respective determination of whether there was a new transaction in each of the plurality of data time periods for each respective online login identification into the trained predictive model assigned to the online login identification;
retrieving data for a plurality of the online login identifications via an online connection in an order of the probability scores from higher probability of having a new transaction to lower probability of having a new transaction since the last successful data retrieval time.

11. The system of claim 10 , wherein the process training a plurality of predictive models further comprises:
sorting the online login identifications from higher probability based on the probability scores from higher probability of having a new transaction to lower probability of having a new transaction since the last successful data retrieval time; and
wherein retrieving data for a plurality of the online login identifications is based on the sorted online login identifications.

12. The system of claim 10 , wherein the data time periods are days and the plurality of data time periods is a plurality of consecutive days.

13. The system of claim 12 , wherein the plurality of consecutive days is at least 30 days.

14. The system of claim 10 , wherein the plurality of consecutive days is at least 90 days.

15. The system of claim 10 , wherein the learning algorithms are each random forest machine learning algorithms.

16. The system of claim 10 , wherein retrieving data for a plurality of the online login identifications is performed for a predetermined percentage of the online login identifications.

17. The system of claim 16 , wherein the predetermined percentage is selected from the group consisting of 90%, 80%, 70%, 60%, 50%, 40%, 30%, 2

Assignees

• Intuit Inc

Inventors

• Misra Siddhartha
• Rabbani Tarek
• Cherukuri Nageswara Prasad
• Nene Kishore
• Kumari Neha
• Kakirwar Prateek Vasant
• Subramanian Mahesh

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

• G06F21/316

  by observing the pattern of computer usage, e.g. typical user behaviour · CPC title

• G06Q40/06

  Asset management; Financial planning or analysis · CPC title