Airline Empires - Airline and Aviation Articles, Airline Simulation - New Methods to Forecast Passenger Demand. Part 1: Page 4

New Methods to Forecast Passenger Demand. Part 1

Page 4 of 4

This research was designed to show both the accuracy that neural networks can provide to market forecasting in non-stop airline markets, and to identify the factors that determine the profitability of these markets. Five routes will be chosen with an attempt to simulate different scenarios with each: Atlanta Hartsfield-Jackson International (ATL) – Los Angeles International (LAX), Dallas – Ft. Worth International (DFW) – Chicago O’Hare International (ORD), Orlando International (MCO) – New York JFK International (JFK), New York Laguardia (LGA) – Washington Reagan National (DCA), and Atlanta Hartsfield – Jackson International (ATL) – Orlando International (MCO). The ATL-LAX route is selected because Delta Airlines has had a monopoly on origin-and-destination traffic on this route for several years. It will provide information for a monopolistic route that can be applied, with some liberty, to other monopolistic routes. DFW-ORD has been a major route for competition among the legacy carriers. With Dallas being a major American hub, O’Hare being a major United hub, and DFW being a minor Delta hub until recently, this route is well suited to identifying variables related to this type of competition. Since JetBlue has entered the market in JFK, its affect on the east coast Florida market has been substantial.

The MCO-JFK route will be able to include JetBlue’s factor in this route’s profitability, as well as Delta’s competitive answer, Song. Other New York airports, such as LGA, and EWR, will be examined for relationships as well, to retrieve an example of how one market affects another area market. LGA-DCA is well known as business traveler market, and has received recognition on several carriers as “the shuttle.” With hourly flights, this provides insight into a market with extreme frequency and the largest business market in the domestic U.S. ATL-MCO will provide understanding of a long held legacy route that has very successfully been infiltrated by a low cost airline. Both Delta and AirTran have been serving the ATL-MCO market with different sized equipment for several years, which should give some insight into low cost carrier entry and aircraft size on a market.

The independent variables that will be entered into the neural network must be as exhaustive a set as can be allowed. This will be crucial in determining which factors have a strong effect, and which have no effect whatsoever on the market.

Independent Variables:

Average Ticket Price for Market
Average Ticket Price for Non-Stop Market
Frequency
Seats per Aircraft
Total Seats
Competing Markets: -Average Ticket Price for Market -Average Ticket Price for Non-Stop Market -Frequency -Seats per Aircraft -Total Seats
Season (Quarter)
Total Number of Carriers

Each variable will be tested for correlation with significance at the 95% confidence level. Only those variables which have proven to have a significant correlation will be used in each market’s forecast. Historical data from 1993 to 1999 will be entered, and the neural network will be trained using this data in the AiNet v1.1 neural network software. The first prediction will be for Q1, 2000, and predictions will continue until Q3, 2005. After each prediction, the neural network will be retrained using actual data for that period. For instance, once the neural network predicts Q1, 2000, actual data from Q1, 2000 will be used to predict Q2, 2000. This will ensure the resultant data is relying on actual data rather than forecasted data, which should limit compounding error. Mean absolute error will be calculated for each forecast, and reported in the final results.

Delicious

In the next part of this article, we'll take a look at the results and analyze whether or not neural networks can successfully be applied to market forecasts.

Continued with Part 2
References.

Write Comment
Name:	Guest
Title:
Comment:

Comments

References
Written by Guest on 2006-09-18 11:02:02

Is there a list of references for this?

Where is the rest?
Written by Guest on 2006-09-18 10:56:15

So where is part 2?

feed back
Written by Guest on 2006-08-27 03:42:37

Actually there are many methods to forecast,
it depands about what the analysis needs
in airline we have to look to short term, medium terms, and long terms objective from forecasting so
we seasonilty, we have trends and econometric approaches
hope to be clear
Mohammed Salem
www.freewebs.com/wingsofwisdom/

Related Items:

<< Start < Previous 1 2 3 4 Next > End >>