來源:
Kate Smith-Miles
<ijcnn2012@ieee-cis.org>
收信: 96356511@nccu.edu.tw , tsaih@mis.nccu.edu.tw
標題: IJCNN 2012 Paper #462 Decision Notification
日期: 21 Feb 2012 18:34:18 -0000
Dear Author(s),
Congratulations! On behalf of the IJCNN 2012 Program Chairs, we are pleased to inform you that your paper:
Paper ID: 462
Author(s): Shin-Ying Huang and Rua-Huan Tsaih
Title: The Prediction Approach with Growing Hierarchical Self-Organizing Map has been accepted for presentation at the 2012 International Joint Conference on Neural Networks and for publication in the conference proceedings published by IEEE. This email provides you with all the information required to complete your paper and submit it for inclusion in the proceedings. A notification of the presentation format (oral or poster) and timing of that presentation will be sent in a subsequent email.
Here are the steps:
1. Please address the attached REVIEWERS' COMMENTS which are intended to improve the final manuscript. Final acceptance is conditional on appropriate response to the requirements and comments.
2. Please prepare your manuscript for final camera ready submission by
following the formats described on the conference web site and using the IEEE templates:
http://www.ieee-wcci2012.org
Once you are ready to submit it, please go to:
http://ieee-cis.org/conferences/ijcnn2012/upload.php?PaperID=462
to submit your final camera-ready paper. On this page you will need to use the following password:
hb8z8h49f
Please do adhere the strict deadline for final manuscript submission April 2, 2012. Any papers submitted after this date will not be included in the
proceedings. The paper must be re-submitted even if the reviewers indicated that no changes are required.
3. In order for your paper to be published in the conference
proceedings, a *signed IEEE Copyright Form* must be submitted for each paper. IJCNN 2012 has registered to use the IEEE Electronic Copyright (eCF) service. The confirmation page shown after submitting
your final paper contains a button linking directly to a secure IEEE eCF site which allows electronic completion of the copyright assignment process. In case it fails, please have the completed IEEE Copyright
Form, found at http://www.ieee.org/web/publications/rights/copyrightmain.html, emailed it to Publication Co-Chair, Daryl Essam (d.essam@adfa.edu.au).
IMPORTANT: No paper can be published in the proceedings without being accompanied by a Completed IEEE Copyright Transfer Form. You must complete and submit this form to have your paper included in the conference proceedings.
4. Register for the conference at http://www.ieee-wcci2012.org by clicking on the conference registration link on the right-hand side of the main page.
IMPORTANT: Each paper must have a corresponding registered author to be included in the proceedings. Papers that do not have an associated registered author will not be included in the proceedings. The deadline for author registration is April 2, 2012 so be sure to register by that time to ensure that your paper is included in the proceedings. Please ensure that you complete your registration early.
If you have any questions regarding the reviews of your paper please contact Kate Smith-Miles <ijcnn2012@ieee-cis.org>.
Sincerely, Kate Smith-Miles <ijcnn2012@ieee-cis.org>
REVIEWERS' COMMENTS
--- REVIEW NO. 1
Originality: Weak Accept Significance of topic: Accept Technical quality: Weak Accept
Relevance to IJCNN 2012: Weak Accept Presentation: Accept
Overall rating: Weak Accept
Reviewer's expertise on the topic: High Suggested form of presentation: Poster Best Paper Award nomination: No Comments to the authors:
The study consists of an application of the Growing Hierarchical Self-organizing Map to the problem of
financial statement fraud detection. The issue addressed is of clear practical importance, at to date, there
have been few (if any) applications of GHSOM for this task.
The paper is quite clearly-written and is fairly easy to follow.
The primary drawback of the paper is that the results of the study are not benchmarked against any other
method (obviously many traditional methods of classification exist). Therefore, the reader cannot form any
real view as to whether the method applied actually produces good in/out sample results.
Consequently,
the paper looks to be more of poster standard rather than oral presentation standard.
--- REVIEW NO. 2
Originality: Neutral
Significance of topic: Neutral Technical quality: Weak Reject Relevance to IJCNN 2012: Accept Presentation: Weak Reject
Overall rating: Neutral
Reviewer's expertise on the topic: Medium Suggested form of presentation: Poster Best Paper Award nomination: No Comments to the authors:
Contributions
In this paper, the Growing Hierarchical Self-Organizing Map (GHSOM) was applied as a prediction tool for financial fraud detection (FFD), based on the
assumption that there was a certain spatial relationship amongst fraud and non- fraud samples. By comparing the results based on training samples and testing samples, the authors concluded that the prediction performance was
acceptable.
Positive aspects
This is a new application of GHSOM to FFD, though the advantages of proposed method have not been well demonstrated by the results.
Observed deficiencies and suggestions
The entire study in this paper was based on the assumption that there was a certain spatial relationship amongst fraud and non-fraud samples.
However, there was a lack of theoretical foundation for such assumption. A citation should be provided if such assumption has been applied in literature, or a further explanation should be given.
The advantage of GHSOM over SOM was mentioned in the Introduction.
Unfortunately it was not tested or demonstrated by the experiment results.
Lack of supporting evidences for final results, why error lower than
20% is considered to be acceptable? Comparing to whom? Or based on what criterion?
Was the parameter beta adjusted based on the classification error? If yes, whether or how this parameter can be adapted in new datasets?
Any limitation of the proposed approach?
There was no definition for FT and NFT. There was no definition for Type I and Type II error.
The conclusion repeats some content already used.
Repeated typo. "spacial" should be "spatial"?
--- REVIEW NO. 3
Originality: Accept
Significance of topic: Weak Accept Technical quality: Accept
Relevance to IJCNN 2012: Accept Presentation: Accept
Overall rating: Weak Accept
Reviewer's expertise on the topic: Low Suggested form of presentation: Any Best Paper Award nomination: No Comments to the authors:
The paper goes into great detail of the methodology. My main criticism is that the cost of a type I error and a type II error are different, so summing the two is not optimal. It is not clear whether the method is superior to existing methods of identifying fraud, because no comparison is made.