Airline demand forecasting utilizing unobscuring and unconstraining

• US 10,489,871 B2
• Filed: 07/19/2017
• Issued: 11/26/2019
• Est. Priority Date: 03/08/2013
• Status: Active Grant

• Current Assignee: American Airlines Incorporated (American Airlines Group Incorporated)
• Original Assignee: American Airlines Incorporated (American Airlines Group Incorporated)
• Inventors: Tracy, Jonathan, Fox, James, Bowles, Julianne, Costello, Michael, Muzich, Daniel, Trenga, Thomas
• Primary Examiner(s): Allen, Akiba K
• Application Number:
  US15/654,240
• Publication Number:
  US 20170316536A1
• Time in Patent Office: 860 Days
• Field of Search: 705 11, 705 5, 705 731, 705 711, 705 712
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: G06Q 50/40 Business processes related to the transportation industry shipping G06Q10/083

Abstract

A computer-based system for unobscuring and/or unconstraining demand is disclosed. Via use of the system, actual airline seat bookings may be restated in an unobscured and/or unconstrained form, facilitating improved demand forecasts for subsequent seat bookings. In this manner, seat protects may be better allocated to align with actual demand, thus increasing revenue.

Claims

1. method for unobscuring a demand level for airline seats, the method comprising:

receiving, at a first computing system comprising a first processor configured for unobscuring demand for airline seats, a bookings table representing actual seat bookings for a flight, wherein the actual seat bookings are grouped into a plurality of fare classes;

determining, by the first processor, which fare classes in the plurality of fare classes have demand that is obscured due to the availability of seats for booking in a lower fare class in the plurality of fare classes;

calculating, by the first processor and via execution of an expectation management (EM) algorithm, the unobscured demand for at least one fare class in the plurality of fare classes;

converting, by the first processor, the unobscured demand for the at least one fare class into integer values representing seat bookings in the respective fare classes;

updating, by the first processor, the bookings table with the integer values to form an unobscured bookings table; and

transferring, by the first computing system and to a second computing system configured with a demand forecasting system operative thereon, the unobscured bookings table,

wherein the second computing system utilizes the unobscured bookings table to generate an unobscured demand forecast for the flight,

wherein the second computing system transfers, to a third computing system configured for taking seat bookings, the unobscured demand forecast, and

wherein the third computing system, based on the unobscured demand forecast, opens an additional fare class on the flight for seat bookings.

2. The method of claim 1, wherein the third computing system takes a seat booking in the additional fare class via a public network.

3. The method of claim 1, wherein the plurality of fare classes are configured with parent-child relationships, and

wherein, in the calculating the unobscured demand:

the unobscured demand for a child class is used as the upper bound for the EM algorithm as applied to the parent class,

the top-down bookings for the class in question are used as the lower bound for the EM algorithm as applied to the parent class, and

the EM algorithm is iteratively executed by the processor until a convergent solution for unobscured demand is obtained for the at least one fare class.

4. The method of claim 1, wherein the EM algorithm is iterated at least 10 times to obtain a convergent solution for unobscured demand.

5. The method of claim 1, further comprising combining, by the first processor, the unobscured bookings table with a historical booking table to form a revised historical bookings table.

6. The method of claim 1, wherein the unobscured demand forecast is used by at least one of a reservation system or a revenue management system to determine whether to open an additional fare class for bookings.

7. The method of claim 1, wherein the determining the fare classes for which demand is obscured further comprises:

converting, by the first processor, the bookings table into a top-down format;

identifying, by the first processor, the lowest fare class for which bookings were received in a first time period;

identifying as an obscured class, by the first processor, the parent class of the lowest fare class for which bookings were received; and

identifying as an obscured class, by the first processor, all fare classes in the plurality of fare classes that are higher than the parent class of the lowest fare class for which bookings were received.

8. The method of claim 7, wherein the EM algorithm comprises:

expBkgs=mean+(stdev*stdev)*(pdfLower−

pdfUpper)/(cdfUpper−

cdfLower);

where;

expBkgs is the expectation of seat bookings in the current fare class;

i.e. the unobscured bookings;

upperBound is the upper bound of the unobscured seat bookings in the current fare class;

lowerBound is the lower bound of the unobscured bookings in the current class;

mean is the mean of the seat bookings for the current fare class in a cohort to which flight belongs;

stdev is the standard deviation of the bookings for the current fare class in a cohort to which the flight belongs;

cdfUpper is the cumulative distribution function of upperBound;

cdfLower is the cumulative distribution function of lowerBound;

pdfUpper is the probability density function of upperBound; and

pdfLower is the probability density function of lowerBound.

9. The method of claim 1, wherein the calculating the unobscured demand for the at least one fare class is terminated upon determining, by the first processor, that the unobscured demand for the current class rounds to zero.

10. The method of claim 1, further comprising weighting, by the first processor, the unobscured demand for the at least one fare class by a weighting parameter.

11. The method of claim 10, wherein the weighting parameter is determined based at least partially on the departure date of the flight.

12. The method of claim 10, wherein the weighting parameter is determined based at least partially on the departure date of the flight being a holiday or a weekend.

13. The method of claim 1, further comprising

accessing, by the first processor, the bookings table, wherein the bookings table further contains forecasted demand for the flight, and wherein at least one fare class in the plurality of fare classes is closed and contains no bookings;

determining, by the first processor and using the seat bookings and the forecasted demand, the monetary stimulation value of opening the closed class;

determining, by the first processor and using the seat bookings and the forecasted demand, the monetary dilution penalty of opening the closed class; and

transferring, by the first computing system and over the first network, the monetary stimulation value and the monetary dilution penalty to the second computing system to generate booking instructions.

14. The method of claim 1, further comprising:

accessing, by the first processor, the bookings table, wherein the bookings table further contains information representing a net prime class seat booking for a flight, and wherein the net prime class seat booking occurred during a first time period;

determining, by the first processor, a floor class for the net prime class seat booking in the bookings table;

determining, by the first processor, a displacement class for the net prime class seat booking in the bookings table;

determining, by the first processor, a target class for the net prime class seat booking in the bookings table, wherein the target class is the higher of the floor class and the displacement class; and

remapping, by the first processor, the net prime class seat booking in the bookings table into its target class to form a remapped bookings table.

15. The method of claim 1, further comprising:

accessing, by a second processor configured for unconstraining demand for airline seats, the second processor forming part of the first computing system, the unobscured bookings table;

determining, by the second processor, the fare classes in the plurality of fare classes for which demand is constrained;

calculating, by the second processor, the unconstrained demand for each constrained fare class in the plurality of fare classes;

converting, by the second processor, the unconstrained demand for each fare class into integer values representing seat bookings in the respective fare classes;

updating, by the second processor, the unobscured bookings table with the integer values to form an unconstrained bookings table;

transferring, by the first computing system and to the second computing system, the unconstrained bookings table;

utilizing, by the second computing system, the unconstrained bookings table to generate an unconstrained demand forecast for the flight;

transferring, by the second computing system and to the third computing system, the unconstrained demand forecast; and

opening, based on the unconstrained demand forecast and by the third computing system an additional fare class on the flight for seat bookings.

16. The method of claim 15, further comprising taking, by the third computing system and over a public network, a seat booking in the additional fare class.

17. The method of claim 15, wherein the determining the fare classes for which demand is constrained comprises identifying the fare classes that were opened for bookings during a particular time interval.

18. The method of claim 15, further comprising weighting, by the second processor, the unobscured demand for the at least one fare class by a weighting parameter.

19. The method of claim 18, wherein the weighting parameter is determined based at least partially on the departure date of the flight.