The Job almost Done, it is a forecasting study concerning some airports using signal tracking approach and coefficient of Determination. Yes the result is outstanding just take a look!!!
2. The Job Almost Done
To Achieve less than 1 % Error
By Mohammed Salem Awad
consultant
When I started my carriers I faced many challenges, challenge in
implementing theory in practice, one of these were setting goals and
targets, which basically relay on forecasting, at that time I worked
hard to develop a seasonality model especially for airline, it take
about six month to solve such seasonality model using AREMA
model with the final years student of Aden University, and a paper
about that work was published in 1998 - Warangal conference – India, Yes it was hard and touch paper
we can refer to the following link more details
http://www.slideshare.net/fullscreen/wings_of_wisdom/a-
multiplicative-time-series-model/1 .
The path of that trails followed by a new study concerning
US carriers – developing what so called Fair – Poor Accuracy
Matrix defining a new performance factor as Signal Tracking
and it conjugate impact with coefficient of determination R
to align the best results of line fitting for seasonality model,
and that can evaluate by addressing the displacement and
directional factors of the mathematical model which is also
define 4 regions to hold the right decision as Fair, Mislead,
Unrelated and poor. A detail work we find it in a paper
concerning Major US Airports in the following link
http://www.slideshare.net/fullscreen/Aviation_Articles/accuracy-of-forecasting-model-us-carriers/1
Goals and Targets:
Top managements always ask about achieving goals and targets, but at what level, and what is our
objectives, is it short term targets or long term targets, how we interpolate the results, is it logic to
accept the results or just to implement the formula. Really all this inquires lead us to practice the term
“Forecasting By Objective”
Based on what the analysis proposed there are many methods as
1- Forecasting based R square Value ( Best Value )
2- Forecasting based On Setting Signal Tracking to Zero
3- Forecasting based On Setting Signal Tracking in accepted region ( -4 < S.T. < 4 )
4- Forecasting based on most recent year
5- Forecasting to meet a specified target ( Trend Target ).
We will address the first 3 methods and we will compare the results
3. Forecasting
Accuracy Matrix:
One of the new creative methodology. It
basically developed based on two main
estimated mathematical parameters,
Displacement and Directional factors
which has a consequence impacts on R
and Signal Tracking by setting boundary
accuracy:
For Fair forecasting, the model should
fulfill these criteria – (Golden Rules)
R2
≥ 80 and
Signal Tracking should be - 4 ≤ S. T. ≤ + 4
Then by developed Forecasting Accuracy Matrix the following outcomes will be concluded
1- Fair Forecast – when R2
and Signal Tracking are in bond.
2- Mislead – Displacement Issue. This case when
R2
is in bond and Signal Tracking is out bond.
we can adjusted signal tracking to be in bond
when there is a room for R2
in the same
analysis so that it can be consider as a fair
forecast.
3- Unrelated – Directional Issue. This case when
R2
is out of the bond and Signal Tracking in the
bond. i.e the balance of accumulated error
without any correlation
4- Poor Forecast – when both R2
and Signal Tracking are out of the bond ( Total Mess).
This matrix manipulate the four decision regions to develop the right and best picture of the
accuracy of forecasting. And to enhance the process of decision making for airline data analysis
especially traffic forecasting, that maps the overall forecasting accuracy of Airports.
4. Data Base ( 2010 – 2012 )
First Analysis:
1- Forecasting based R square Value ( Classical Method )
2- Forecasting based On Setting Signal Tracking to Zero
3- Forecasting based On Setting Signal Tracking in accepted region ( -4 < S.T. < 4 )
4- Forecasting based on most recent year
6. Third Analysis:
3- Forecasting based On Setting Signal Tracking
in accepted region ( -4 < S.T. < 4 ) ( Max/Min S. T. approach )
7. The Results
( Third Analysis):
The accuracy of forecasting in third one is almost fair, yes the signals tracking are in the bond
and the coefficient of Determination is high ( 93.6 % ). While for comparison propose, we
evaluate the errors of 2013 ( 4 months actual data with the forecast ), the results listed below
with errors less than one.
The four months of 2013
comparing result shows that the
Error is less than 1 % !
8. Get Your Own Targets of ( 2013- 2014)
By Filling the Forms : The first 50 requests are Free !!!!!!!!
To: smartdecision2002@yahoo.com ( Mohammed S. Awad )
(Monthly - Form)
Months 2010 2011 2012
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
(Annually - Form)
Years Data
2005
2006
2007
2008
2009
2010
2011
2012