Join our Newsletter
Contact
webmaster: Sven F. Crone
Centre for Forecasting
Lancaster University
Management School
Lancaster LA1 4YF
United Kingdom
Tel +44.1524.592991
Fax +44.1524.844885
eMail sven dot crone (at)
neural-forecasting dot com
ISF2005 Neural Network Forecasting Competition
The organisers would like to thank everyone for participating in the 2005 ISF neural network forecasting competition. We are thankfull for the many suggestions on how to improve the experimental design and are planning an extended forecasting competition for the 2006 ISF.
Results
|
MdAPE
|
Rank |
MAPE
|
Rank |
MAE
|
Rank |
RMSE
|
Rank | Sum of Ranks | Total Rank | ||
| 1 | Pant |
8,4%
|
1
|
10,3%
|
1
|
383,9
|
1
|
826,31
|
6
|
3 | 1 |
| 2 | Yan |
11,5%
|
3
|
12,0%
|
2
|
463,7
|
2
|
727,61
|
5
|
7 | 2 |
| 3 | Dhawan |
10,4%
|
2
|
12,0%
|
3
|
494,0
|
3
|
552,66
|
1
|
8 | 3 |
| 4 | Merkwirth |
11,8%
|
4
|
12,2%
|
4
|
497,8
|
4
|
697,86
|
3
|
12 | 4 |
| 5 | Da Silva |
12,5%
|
5
|
14,3%
|
6
|
568,0
|
6
|
884,45
|
7
|
17 | 6 |
| 6 | undisclosed |
13,2%
|
6
|
13,1%
|
5
|
517,8
|
5
|
711,35
|
4
|
16 | 5 |
| 7 | undisclosed |
15,3%
|
7
|
15,4%
|
7
|
608,5
|
7
|
1168,40
|
9
|
21 | 7 |
| 8 | undisclosed |
17,0%
|
8
|
16,0%
|
8
|
633,4
|
8
|
672,76
|
2
|
24 | 8 |
| 9 | undisclosed |
17,4%
|
9
|
18,6%
|
9
|
756,4
|
9
|
913,13
|
8
|
27 | 9 |
| 10. | not accepted | ||||||||||
| 11. | not accepted | ||||||||||
| 12. | not accepted | ||||||||||
| 13. | not accepted | ||||||||||
| 14. | not accepted |
Upon request, only the top 5 submissions were identified by author. All others are disclosed upon request by the participants. 5 submissions were not accepted.
Datasets
The following page will host the datasets for the ISF 2005 time series competition for forecasting principles.
Dataset 1 is already published, as it represents a well known & researched dataset.
Dataset 2 is undisclosed.
 | Please download the datasets and the corresponding questionnaires as a single MS Excel 2003 File - to save right click button and "save target as" |
Dataset 1: Airline passenger Data
| Name: | airline |  |
| Description: | International airline passengers | |
| Origin | Brown (1956), Box & Jenkins (1976) | |
| Frequency: | monthly totals in thousands | |
| Observations | Jan 1949 - Dec 1956, extended to Dec 1960 | |
| Predictions | Jan 1961 - Dez 1964 | |
| References | S. G. Makridakis, S. C. Wheelwright, and R. J. Hyndman, Forecasting: methods and applications. New York: Wiley, 1998. Original: Brown (1956), Box & Jenkins (1976), O.D. Anderson (1976) and O'Donovan(1983). |
Dataset 2: Unknown Time Series
| Name: | undisclosed |  |
| Description: | undisclosed | |
| Origin: | undisclosed | |
| Frequency: | monthly totals in thousands | |
| Observations | Jan 1900 (date changed) - Jan 1907 | |
| Predictions | Feb 1907 - Jun 1910 | |
| References | undisclosed |
Submission Instructions
NOTE: These instructions are taken from the MS Excel file contaiing the datasets.
Please download the file on these pages and follow the detailed instructions given therein.
INDIVIDUAL DEADLINE EXTENSIONS ARE AVAILABLE UPON REQUEST !!!
1. Submit your abstract (200 words, no maths etc.) summarizing your neural network method & modelling process
- Where to submit: ISF05 website http://www.isf2005.org/submissions.php
- Follow the instructions on the webpage!!! - SUBJECT AREA "Neural Nets"
- When to submit: deadline is 28 February 2005
3. Fill in your forecast results for 2 time series on sheets "Dataset 1" and "Dataset 2" in the MS Excel file
- input your solutions for the airline time series on sheet "Dataset 1" replacing the ?s in column C (C146:C193)- input your solutions for the unknown time series on sheet "Dataset 2" replacing the ?s in column C (C87:C127)
- as the original data is given in thousands, please round to integer values.
4. Fill in your description of your forecasting method on "Questionnaire 1" and "Questionnaire 2" in the MS Excel file
- please give a detailed description of your method / neural network architecture used and process of determining it- use a verbal or mathematical description (formatting is not a relevant issue - content is!)
- it should allow someone unfamiliar with your modelling approach to recapture the relevant decisions to generate your forecasts
3. Submit your description of your forecasting method via email
- rename this file "ISF05_NN_yournamehere.xls", replacing "yournamehere" with the first author family name.- send this file in an email to ISF05-NN-competition@neural-forecasting.com with "ISF05-NN-competition" in the subject
- include your name, filename and email-address in your email body
- When to submit: deadline is
- Should you not receive a confirmation within 48 hours of your submission contact s.crone@neural-forecasting.com
--> Notification of acceptance of abstracts: 31 March 2005