o d passenger demand forecasting n.
Download
Skip this Video
Loading SlideShow in 5 Seconds..
O&D Passenger Demand Forecasting PowerPoint Presentation
Download Presentation
O&D Passenger Demand Forecasting

Loading in 2 Seconds...

play fullscreen
1 / 11

O&D Passenger Demand Forecasting - PowerPoint PPT Presentation


  • 154 Views
  • Uploaded on

O&D Passenger Demand Forecasting. John Blankenbaker, Wassim Chaar DT Operations Research AGIFORS Reservations and YM 2003 Conference June 2 - 5, 2003 Honolulu, Hawaii. Forecasting Methods. Bottom-Up (Direct) Methods Independent time-series Correlated time-series models Top-Down Methods

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about 'O&D Passenger Demand Forecasting' - chickoa


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
o d passenger demand forecasting

O&D Passenger Demand Forecasting

John Blankenbaker, Wassim Chaar

DT Operations Research

AGIFORS Reservations and YM 2003 Conference

June 2 - 5, 2003

Honolulu, Hawaii

forecasting methods
Forecasting Methods
  • Bottom-Up (Direct) Methods
    • Independent time-series
    • Correlated time-series models
  • Top-Down Methods
    • Forecast Enrichment
    • Forecast Inference (FI)
fi method
FI Method
  • Forecast unconstrained passenger demand at the leg cabin level.
  • Infer unconstrained passenger demand at O&D itinerary class level.
  • Aggregate O&D itinerary class level forecast to leg bucket level for use in leg optimization.
advantages of fi
Advantages of FI
  • Leg cabin forecasts are more accurate and error propagation can be controlled.
  • Sophisticated inference engine used to derive O&D itinerary class level forecasts.
  • Provides “natural” O&D controls by adjusting passengers mix
  • No adjustment needed to produce leg bucket level forecasts.
forecast accuracy experiment
Forecast Accuracy Experiment
  • Used revenue accounting data.
  • Usedmean absolute error (MAE) on nests that were open at departure.
  • Used weekly checkpoints from 63 days to 7 days before departure.
  • Considered the Y01-Y04 nest, the Y01-Y08 nest, the Y01-Y12 nest and the Y cabin total.
experimental results example 1
Experimental Results: Example 1

Mean Absolute Error Y04 Nest

Mean Absolute Error Y08 Nest

(N = 2701)

(N = 2099)

LEG

LEG

FI

FI

Mean Absolute Error

Mean Absolute Error

063

056

049

042

035

028

021

014

007

063

056

049

042

035

028

021

014

007

Check Point

Check Point

Mean Absolute Error Y12 Nest

Mean Absolute Error Y16 Nest

(N = 1554)

(N = 933)

LEG

LEG

FI

FI

Mean Absolute Error

Mean Absolute Error

063

056

049

042

035

028

021

014

007

063

056

049

042

035

028

021

014

007

Check Point

Check Point

experimental results example 2
Experimental Results:Example 2

Mean Absolute Error Y04 Nest

Mean Absolute Error Y08 Nest

(N = 88)

(N = 73)

LEG

LEG

FI

FI

Mean Absolute Error

Mean Absolute Error

063

056

049

042

035

028

021

014

007

063

056

049

042

035

028

021

014

007

Check Point

Check Point

Mean Absolute Error Y12 Nest

Mean Absolute Error Y16 Nest

(N = 65)

(N = 35)

LEG

LEG

FI

FI

Mean Absolute Error

Mean Absolute Error

063

056

049

042

035

028

021

014

007

063

056

049

042

035

028

021

014

007

Check Point

Check Point

experimental results example 3
Experimental Results: Example 3

Mean Absolute Error Y04 Nest

Mean Absolute Error Y08 Nest

(N = 15)

(N = 15)

LEG

LEG

FI

FI

Mean Absolute Error

Mean Absolute Error

063

056

049

042

035

028

021

014

007

063

056

049

042

035

028

021

014

007

Check Point

Check Point

Mean Absolute Error Y12 Nest

Mean Absolute Error Y16 Nest

(N = 11)

(N = 8)

LEG

LEG

FI

FI

Mean Absolute Error

Mean Absolute Error

063

056

049

042

035

028

021

014

007

063

056

049

042

035

028

021

014

007

Check Point

Check Point

experimental results example 4
Experimental Results:Example 4

Mean Absolute Error Y04 Nest

Mean Absolute Error Y08 Nest

(N = 11)

(N = 7)

LEG

LEG

FI

FI

Mean Absolute Error

Mean Absolute Error

063

056

049

042

035

028

021

014

007

063

056

049

042

035

028

021

014

007

Check Point

Check Point

Mean Absolute Error Y12 Nest

Mean Absolute Error Y16 Nest

(N = 4)

(N = 4)

LEG

LEG

FI

FI

Mean Absolute Error

Mean Absolute Error

063

056

049

042

035

028

021

014

007

063

056

049

042

035

028

021

014

007

Check Point

Check Point

summary
Summary
  • Despite input data problems, FI generated O&D itinerary class forecasts which were, when aggregated to the leg bucket level, quite good.
  • Our prototype is available for experimentation and benchmarking purposes.