Connectionists: BRICS-CCI & CBIC 2013 Data Mining Competition Announcement ("Porto de Galinhas" - Brazil)

Tue Jun 18 22:10:06 EDT 2013

Dear All,

The organizers of 1st BRICS and 11th Brazilian Congresses on Computational
Intelligence <http://brics-cci.org/> are proud to announce an
Algorithm Competition,
to be held in the beach village of "Porto de Galinhas", Brazil on 8-11th
September. We would be very pleased if you could encourage your colleagues
and students to enter the competition and also if you could publicize this
much interesting co-located activity of our Congress. The submissions for
main tracks of BRICS-CCI and CBIC 2013 are now closed with a record of over
350 submissions, *but *Competitions (i.e. SW as well as
Theses/Dissertations) and other Symposia are still open for submissions.

We have brought this year competition topic from market demands on the
traditional credit risk assessment domain, but under an innovative
perspective, as previously done by us in the PAKDD competitions 2009 &
2010. This endeavor was not possible without NeuroTech
S.A.<http://www.neurotech.com.br/english.html>
 (co-organizer) so that we are quite happy to host the first data mining
competition in Brazil. -We do hope to see you next September! Congress is
packed with interesting speakers (
http://brics-cci.org/confirmed-invited-speakers-by-october-2012/)

*Competition Deadline:* July, 31st 2013

*Overview:*

-The competition is open for academia and industry, accessible either
through the BRICS-CCI & CBIC 2013 Conference site (
brics-cci.org/ci-algorithms-competition-ciac) or directly to the
competition server (brics-cci.neurotech.com.br).

-This year's Competition, is on the well known application of credit
scoring, focused on the effects of temporal degradation of performance and
seasonality of payment delinquency.

-There is a real-time LeaderBoard for stimulating the competitors' daily
participation and to allow some parameter adjustment.

*Problem Summary: **Credit Risk Assessment System Robustness Against
Degradation and Seasonal Variation*

The offer of credit for potential clients is a very important service for
stimulating consumption in the market. Despite being among the oldest
application domains for data mining, there are some difficulties related to
credit scoring which are often overlooked by modelers, namely:

·         In general, there are only data about the company's clients for
modeling, but not about the rejected applicants thus representing a
strongly biased sample of the market given that a systematic procedure
focused on the problem target (payment default) has been applied for their
selection.

·         The data is also collected from a time interval in the past for
developing a model to be applied in a future time. Despite the absence of
any drastic change in the economy, gradual market changes and geographical
expansion occur affecting the performance of the model estimated on the
modeling data set.

·         Added to the degradation, there is a seasonality effect that
causes variations on the monthly payment delinquency.

 This competition focuses on two features of the credit risk assessment
model that are presented as tasks for the participants:

*Task 1*: Robustness against performance degradation caused by market
gradual changes along few years of business operation.

*Task 2*: Fitting of the estimated delinquency produced by the estimation
model to that observed on the actual data for the applications approved by
the model for Task 1.

-Participants will download a labeled data set from a two-year period
(2009-2010) for modeling and will submit scores to the test dataset from a
year ahead (2011). The leaderboard results will be assessed on almost than
10% of the records from the test dataset while the whole test dataset will
be used for the final competition performance evaluation. These data sets
come from the private label credit card operation of a major Brazilian
retail chain, along stable inflation condition (2009-2011).

-Competitors may use any modeling technique they wish for either task and
the official performance metrics will be the area under the ROC curve for
task-1 and the distance D of the Chi-square statistics for the monthly
delinquency estimated in task-2.

*Webpage: *brics-cci.org/ci-algorithms-competition-ciac/     or
brics-cci.neurotech.com.br

*Contact:* brics2013 at neurotech.com.br

*Competition Organizers:*

-Paulo Jorge Leitão ADEODATO <pjla at cin.ufpe.br> Center of Informatics,
Federal University of Pernambuco / Neurotech S.A., Brazil

-Swagatam Das <swagatam.das at isical.ac.in> Indian Statistical Institute,
India

-Domingos S. P. Salazar <domingos_salazar at hotmail.com> Distance Learning
Unit, Federal Rural University of Pernambuco / Neurotech S.A., Brazil

-- 
<http://fbln.pro.br/images/FBLN_Footer.jpg>
Prof. Fernando Buarque <http://fbln.pro.br/images/FBLN_Footer.jpg>, BSc MSc
DIC PhD Hab., Senior Member IEEE, PQ-2 CNPq
Professor Associado - Escola Politécnica/Universidade de Pernambuco
(POLI<http://www.poli.br/>
/UPE <http://www.poli.br/>)
Coordenador do Mestrado de Engenharia de Computação da UPE
(PPG-EC<http://mestrado.ecomp.poli.br/>
)
Docente do Mestrado de Engenharia de Sistemas da UPE
(PPG-ES<http://ppges.poli.br/>
)
Docente da Graduação de Engenharia de Computação da UPE
(E-Comp<http://www.ecomp.poli.br/>
)
Pesquisador Líder do Grupo de Inteligência Computacional da UPE
(CIRG at UPE<http://www.cirg.ecomp.upe.br/>
)
Assessor de Relações Internacionais da POLI
(ARI at POLI<http://www.poli.br/index.php?option=com_content&view=article&id=690:a-ari&catid=74:informacoes-ari&Itemid=273>
)
Coordenador Técnico do Núcleo de Telemedicina da UPE
(NUTES at UPE<http://www.nutes.upe.br/>
)
Visiting Professor - University of Johannesburg, South Africa (Kingsway
Campus <http://virtualtour.co.za/clients/University/index.html>)
Graduate Faculty - Computer Science/Florida Institute of Technology, USA (
CS at FIT <http://coe.fit.edu/cs/>)
<http://brics-cci.org>
Universidade de Pernambuco / Escola Politécnica de Pernambuco
Rua Benfica, 455 (Bl. 'C' 2. andar) * Bairro: Madalena
CEP 50720-001  *  Recife, Pernambuco - Brasil
Fone: +55(0)81 3184-7542 * Fax: +55(0)81 3184-7581
WEB -> http://www.fbln.pro.br/ * WAP -> http://www.fbln.pro.br/wap.htm

*"Se você quiser educar um homem, comece pela avó dele" (Victor Hugo).*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20130618/1f6794d1/attachment.html>