Applying Logistic Regression to Historical Data to Generate Forecast Coefficients

Abstract

The system applies logistic regression to historical data to generate forecast coefficients. The system creates an updated authorization parameter in real-time iteratively throughout a period of time.

1 . A system comprising:
 a network interface communicating with a memory;   the memory communicating with a processor, wherein the processor communicates with an aircraft unbooked passenger no-show forecast (NSF) engine, a cost engine, an authentication server, an optimizer module and an upgrade analyzer;   the processor, when executing a computer program, executes operations comprising:   communicating, by the processor, with an aircraft central data repository over the internet by invoking logic within modules by passing parameters relating to requests for data;   determining, by the processor, an aircraft next active leg (NAL) of the aircraft passenger based on passenger name record (PNR) characteristics of an aircraft passenger;   analyzing, by the processor, historical show rate data by directional markets and the PNR characteristics;   applying, by the processor, logistic regression to historical PNR data to generate forecast coefficients based on the historical show rate data for a forecast market for the passenger;   aggregating, by the processor, a subset of the forecast coefficients associated with historical passenger data with similar PNR characteristics;   forecasting, by the processor, an NSF for a particular passenger based on the aggregating and based on the PNR characteristics from the aircraft central data repository;   forecasting, by the processor, aircraft flight costs for an aircraft flight;   determining, by the processor and based on the PNR characteristics from the aircraft central data repository, an aircraft booked passenger NSF for each aircraft booked passenger for the aircraft flight, wherein the aircraft booked passenger NSF for each respective aircraft passenger is based upon the aircraft NAL of the respective aircraft passenger;   assigning, by the processor, a forecast market based on the aircraft NAL;   determining, by the processor and based upon the forecast market and PNR characteristics, NSF forecast coefficients for the passenger;   determining, by the processor and within an advanced purchase timeframe, an aircraft unbooked passenger NSF based on the historical show rate data by directional market to predict an unbooked show rate to associate with unbooked seats of the flight,   aggregating, by the processor, the aircraft booked passenger NSF and the aircraft unbooked passenger NSF to create an aircraft flight NSF;   determining, by the processor and using a flight cost prediction model in a forecaster module, an aircraft authorized seat allocation (AU) for the aircraft flight based upon the aircraft NSF to maximize aircraft revenue for the aircraft flight and minimize the aircraft costs of overbooking the aircraft flight,   updating, by the processor and on a database, an authorization parameter to create an updated authorization parameter for the flight based upon the AU;   providing, by the processor and to a revenue management system, the updated authorization parameter,   wherein the revenue management system determines, based on the updated authorization parameter, a number of additional seats to be sold for the flight and a respective price for each additional seat.   
     
     
         2 . The system of  claim 1 , further comprising determining, based upon a first booked passenger itinerary for a first booked passenger, a first booked passenger NAL. 
     
     
         3 . The system of  claim 2 , further comprising:
 determining, based upon the first booked passenger NAL, a first booked passenger forecast market for the first booked passenger; and   determining, based upon the passenger forecast market, a first booked passenger forecast coefficient for the first booked passenger.   
     
     
         4 . The system of  claim 3 , further comprising:
 assigning, based upon the first booked passenger forecast coefficient, a first booked passenger NSF to the first booked passenger; and   linking the first booked passenger NSF to each leg in the first booked passenger itinerary, wherein the flight is one leg in the first booked passenger itinerary   
     
     
         5 . The system of  claim 4 , further comprising:
 providing, by the processor, a certain number of updated tickets to access the flight based on the AU;   providing access information, by the processor and to an airline kiosk, that allows access to the flight to certain passengers with the updated tickets,   wherein the airline kiosk provides an updated boarding passes with machine readable data to the passengers;   receiving, by the processor and from an airport scanner, scanned data from the updated tickets, wherein access to the airplane is provided in response to the airport scanner scanning the machine readable data on the updated tickets and verifying the machine readable data on the updated tickets;   providing denial information, by the processor, that denies access to the flight to denied passengers based on the AU; and   compensating, by the processor, the denied aircraft passengers.   
     
     
         6 . The system of  claim 5 , further comprising:
 storing, by the processor, the data in a file in a database;   tuning, by the processor, the database to optimize database performance,
 wherein the tuning includes placing frequently used files on separate file systems to reduce in and out bottlenecks; 
   designating, by the processor, a key field in data tables to speed searching for the data;   sorting, by the processor, the data according to a known order to simplify the lookup process; and   obtaining, by the processor, the data from the frequently used files.   
     
     
         7 . The system of  claim 6 , wherein the determining the first booked passenger NSF for the flight is dependent upon a previous leg NSF for the first booked passenger, wherein the previous leg NSF is associated with a previous leg, wherein the first booked passenger itinerary is comprised of a plurality of legs, and wherein the flight and the previous leg are among the plurality of legs. 
     
     
         8 . The system of  claim 7 , wherein the first booked passenger NSF is calculated as at least one of a joint probability with the previous leg NSF or a conditional probability given the previous leg NSF. 
     
     
         9 . The system of  claim 8 , further comprising adjusting the first booked passenger NSF in response to determining that the first booked passenger NAL and the flight are scheduled for the same day. 
     
     
         10 . The system of  claim 9 , wherein the determining the booked passenger NSF for each booked passenger comprises analyzing, for each booked passenger, a complete passenger itinerary for the respective passenger. 
     
     
         11 . The system of  claim 10 , wherein each booked passenger NSF is based upon at least one of whether the flight is an originating flight for the respective passenger or whether the flight is a connecting flight for the respective passenger. 
     
     
         12 . The system of  claim 11 , wherein the unbooked seats of the flight is based on (capacity of the flight−number of booked passengers*(booked show rate))/(the unbooked show rate). 
     
     
         13 . The system of  claim 12 , wherein a first booked passenger NSF is based upon at least one of a previous leg, a subsequent leg, a direction of travel of a first booked passenger, whether the flight represents an outbound segment, an intermittent segment or a return segment of a first booked passenger itinerary. 
     
     
         14 . The system of  claim 13 , further comprising:
 applying, by the processor, a default rule to the aircraft flight NSF that is outside of expectations;   determining, by the processor and in response to the default rule, that the forecast coefficient data for the forecast market is not statistically significant; and   utilizing, by the processor, the aircraft unbooked passenger NSF instead of the historical show rate data to revise the aircraft flight NSF.   
     
     
         15 . The system of  claim 14 , wherein the steps are performed in real-time iteratively throughout a period of time. 
     
     
         16 . The system of  claim 15 , further comprising accessing the historical NSF database to determine at least one of the first booked passenger NSF coefficient or the aircraft unbooked passenger NSF, wherein the historical NSF database comprises historical data compiled according to a plurality of factors, the factors comprising forecast market and booking period. 
     
     
         17 . The system of  claim 16 , further comprising:
 determining, based upon the number of booked passengers for the flight and based upon the booked passenger NSF, a number of unbooked passengers for the flight;   determining, based upon a flight date and an analysis date, a number of days before the flight is scheduled to depart; and   determining, based upon the number of days before the flight is scheduled to depart, a booking period, wherein the determining the aircraft unbooked passenger NSF is based upon the booking period.   
     
     
         18 . The system of  claim 17 , wherein the authorized seat allocation minimizes an overbooking cost, wherein the overbooking cost is based upon the flight costs and the flight NSF, and wherein the flight costs comprise a forecasted spoiled seat cost for the flight and a denied boarding cost for the flight. 
     
     
         19 . The system of  claim 18 , further comprising:
 creating, by the processor and in real-time iteratively throughout the period of time, a voucher for the number of additional seats for the flight, wherein the voucher changes throughout the period of time based on the updating the updated authorization parameter and most current conditions to minimize costs; and   providing, by the processor and in real-time iteratively throughout the period of time and to a customer mobile device, the voucher for the number of additional seats for the flight.   
     
     
         20 . A computer-based method, comprising:
 communicating, by a processor, with an aircraft central data repository over the internet by invoking logic within modules by passing parameters relating to requests for data;   determining, by the processor, an aircraft next active leg (NAL) of the aircraft passenger based on passenger name record (PNR) characteristics of an aircraft passenger;   analyzing, by the processor, historical show rate data by directional markets and the PNR characteristics;   applying, by the processor, logistic regression to historical PNR data to generate forecast coefficients based on the historical show rate data for a forecast market for the passenger;   aggregating, by the processor, a subset of the forecast coefficients associated with historical passenger data with similar PNR characteristics;   forecasting, by the processor, an NSF for a particular passenger based on the aggregating and based on the PNR characteristics from the aircraft central data repository;   forecasting, by the processor, aircraft flight costs for an aircraft flight;   determining, by the processor and based on the PNR characteristics from the aircraft central data repository, an aircraft booked passenger NSF for each aircraft booked passenger for the aircraft flight, wherein the aircraft booked passenger NSF for each respective aircraft passenger is based upon the aircraft NAL of the respective aircraft passenger;   assigning, by the processor, a forecast market based on the aircraft NAL;   determining, by the processor and based upon the forecast market and PNR characteristics, NSF forecast coefficients for the passenger;   determining, by the processor and within an advanced purchase timeframe, an aircraft unbooked passenger NSF based on the historical show rate data by directional market to predict an unbooked show rate to associate with unbooked seats of the flight,   aggregating, by the processor, the aircraft booked passenger NSF and the aircraft unbooked passenger NSF to create an aircraft flight NSF;   determining, by the processor and using a flight cost prediction model in a forecaster module, an aircraft authorized seat allocation (AU) for the aircraft flight based upon the aircraft NSF to maximize aircraft revenue for the aircraft flight and minimize the aircraft costs of overbooking the aircraft flight,   updating, by the processor and on a database, an authorization parameter to create an updated authorization parameter for the flight based upon the AU;   providing, by the processor and to a revenue management system, the updated authorization parameter,   wherein the revenue management system determines, based on the updated authorization parameter, a number of additional seats to be sold for the flight and a respective price for each additional seat.