Passenger Recapture Estimation in Airline RM - PowerPoint PPT Presentation

Loading...

PPT – Passenger Recapture Estimation in Airline RM PowerPoint presentation | free to download - id: 1baac2-ZDc1Z



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Passenger Recapture Estimation in Airline RM

Description:

Passenger Recapture Estimation in Airline RM – PowerPoint PPT presentation

Number of Views:39
Avg rating:3.0/5.0
Slides: 22
Provided by: sunn66
Learn more at: http://www.agifors.org
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Passenger Recapture Estimation in Airline RM


1
Passenger Recapture Estimation in Airline RM
  • Shau-Shiang Ja, Beju Rao, Scott Chandler
  • AGIFORS Revenue Management
  • May 8-11 2001

2
Outline
  • Introduction
  • Basic Methodology
  • Preliminary Results
  • Simulation Results
  • Summary
  • Questions/Answers

3
Introduction
  • Demand unconstraining is necessary in order to
    estimate the true demand in a Revenue Management
    (RM) system.
  • RM systems do not always adequately adjust spill
    values to account for the proportion of estimated
    spill that may be recaptured by higher fare
    classes on the same service or by other services.
  • Unconstraining without recapture will result in
    counting the recaptured passengers twice

4
Unconstraining w/o Recapture
  • The two recaptured passengers are counted twice

5
Basic Idea
  • Adjusted Spill Estimated Spill (1-Recapture
    Rate)

6
Basic Idea (cont.)
  • Adjusted Spill Estimated Spill (1-Recapture
    Rate)

7
Related Literature in RM
  • Thomas Gorin (AGIFORS RM Study Group, 2000)
  • There is a sample bias.
  • Sven-Eric Anderssen (International Transactions
    in Operational Research, 1998)
  • Discrete Choice model

8
Basic Methodology
  • Observed DemandY True DemandY RecaptureM2Y
    SpillY
  • Observed DemandY SpillY True DemandY
    RecaptureM2Y

9
Basic Methodology (cont.)
  • Observed DemandY SpillY True DemandY
    RecaptureM2Y
  • RecaptureM2Y Recapture RateM2Y SpillM
  • Therefore,
  • Observed DemandY SpillY True DemandY

  • Recapture RateM2Y SpillM
  • In the above equation, Observed DemandY, SpillY
    and SpillM are known (assuming our unconstraining
    method is perfect) and True DemandY and
    Recapture RateM2Y are unknown.

10
Regression Approach
  • Observed DemandY SpillY True DemandY

  • Recapture RateM2Y SpillM
  • It looks like a standard regression form such
    that both unknowns appear as coefficients.
  • This regression model can be extended to consider
    potential recapture from other lower fare classes
    within the same service and potential recapture
    from other itineraries as well.

11
Regression Model - example
  • Observed Demand2Y Spill2Y b0 b1 Spill2M
    b2 Spill2N

  • b3
    Spill1Y b4 Spill3Y

12
Regression Model (cont.)
  • Observed Demand2Y Spill2Y b0 b1 Spill2M
    b2 Spill2N

  • b3
    Spill1Y b4 Spill3Y
  • The regression parameters (b1, b2, b3, and b4)
    test the influence that related spill has on the
    targeted service/class, while b0 represents the
    true demand.
  • One equation for each service/class/reading_day.
  • One year of spill and observed demand data.
  • At least 40 data points per service/class/reading_
    day.

13
Preliminary Results
14
Preliminary Results (cont.)
  • High spill service/classes have higher observed
    recapture rates.
  • Observed recapture is more noticeable on
    single-leg services.
  • Clumping at origin affects regression results.
  • Occasionally, the regression model returned
    recapture rates greater than one
  • Good data cleaning required on the OD data.
  • Good OD spill estimation is critical.

15
Simulation Model
  • Demand of Y Gamma(a,b) with mean 10
  • Demand of M Gamma(a,b) with mean 15
  • Recapture Rate is 0.3
  • Spill of M Max( Demand of M - Capacity, 0 )
  • Observed Demand of Y Demand of Y 0.3 Spill
    of M
  • Regression Model
  • Observed Demand of Y b0 b1 Spill of M

16
Simulation Result - case 1
  • Demand of Y and Demand of M are independent.

17
Simulation Result - case 2
  • At half of the time (randomly), both Demand of Y
    and Demand of M are increased by 10.

18
Simulation Result - case 3
  • At half of the time (randomly), both Demand of Y
    and Demand of M are multiplied by 2.

19
Future Plan
  • It is possible to integrate the result from
    regression-based approach and QSI-based approach.

20
Summary
  • Unconstraining without recapture results in
    counting the recaptured passengers twice.
  • Recapture rates can be estimated using regression
    from the observed demand and estimated spill on
    all related service/classes.
  • Passenger choice model can be included in the
    calibration of recapture rates in the future.

21
  • Questions?
About PowerShow.com