US 11,657,410 B2
Method, system, and computer program product for wait time estimation using predictive modeling
Richa Garg, Mountain View, CA (US); Walker Carlson, San Diego, CA (US); Varun Sharma, Foster City, CA (US); Nandakumar Kandaloo, Mountain View, CA (US); and Srijoy Aditya, Los Altos, CA (US)
Assigned to Visa International Service Association, San Francisco, CA (US)
Appl. No. 16/954,599
Filed by Visa International Service Association, San Francisco, CA (US)
PCT Filed Dec. 20, 2017, PCT No. PCT/US2017/067530
§ 371(c)(1), (2) Date Jun. 17, 2020,
PCT Pub. No. WO2019/125426, PCT Pub. Date Jun. 27, 2019.
Prior Publication US 2020/0334592 A1, Oct. 22, 2020
Int. Cl. G06Q 30/0202 (2023.01); G06N 20/00 (2019.01); G06Q 10/04 (2023.01); G06Q 20/40 (2012.01); G06Q 30/0207 (2023.01); G06Q 30/0251 (2023.01); G06F 18/214 (2023.01)
CPC G06Q 30/0202 (2013.01) [G06F 18/214 (2023.01); G06N 20/00 (2019.01); G06Q 10/04 (2013.01); G06Q 20/40 (2013.01); G06Q 30/0207 (2013.01); G06Q 30/0251 (2013.01)] 18 Claims
OG exemplary drawing
 
1. A computer-implemented method for generating and applying a predictive wait time estimate using predictive modeling, the method comprising:
receiving, with at least one processor, initial transaction data representative of a plurality of transactions between a plurality of transaction accounts and at least one merchant completed during a sample time period, the initial transaction data comprising, for each transaction of the plurality of transactions, a transaction time and a transaction value;
generating, with at least one processor and based at least partially on the initial transaction data, for each subinterval of a first plurality of subintervals of the sample time period, training data comprising at least one of the following: service rate during the subinterval, number of transactions during the subinterval, total transaction value during the subinterval, mean transaction value during the subinterval, year, month, day of week, or any combination thereof;
generating, with at least one processor and based at least partially on the training data of the sample time period, a predictive model that generates an output of arrival rate for an input comprising at least one time parameter;
generating, with at least one processor and using the predictive model, the predictive wait time estimate for a designated time, wherein generating the predictive wait time estimate comprises:
determining an initial queue length at a start time preceding the designated time;
determining a service rate for each subinterval of a second plurality of subintervals from the start time to the designated time based on transactions processed over time to produce a plurality of service rates;
determining an arrival rate for each subinterval of the second plurality of subintervals from the start time to the designated time using the predictive model to produce a plurality of arrival rates;
determining a difference between an arrival rate of the plurality of arrival rates and a service rate of the plurality of services rates for each subinterval of the second plurality of subintervals, to produce a plurality of changes in queue length for each subinterval of the second plurality of subintervals;
determining a queue length at the designated time based on the initial queue length and the plurality of changes in queue length; and
generating the predictive wait time estimate based on the queue length at the designated time and a service rate at the designated time;
receiving at predetermined intervals, with at least one processor, new transaction data representative of one or more transactions between one or more accounts and the at least one merchant completed during a new sample time period, the new transaction data comprising, for each transaction of the one or more transactions, a transaction time and a transaction value; and
in response to receiving the new transaction data at each predetermined interval of the predetermined intervals:
generating, with at least one processor, new training data based at least partially on the new transaction data;
generating, with at least one processor, a new predictive model based at least partially on the new training data; and
generating, with at least one processor, a new predictive wait time estimate using the new predictive model.