In Brazil, there is a high demand for credit cards. Unfortunately, some people never pay pack their expenses, they just disappear. The objective was to create a model that predicts if somebody won’t pay its debts. The same predictive model can be used to predict defaults on other credit card, loans or any type of credit.

The prediction targets are the clients that do not pay their debt (bad clients). A client is labeled as bad if he made 60 days delay in any payment of the bills contracted along the first year after the credit has been granted. The performance of the predictive model used to compute the final ranking of the teams was measured using area under the receiver operating characteristic curve (AUC) on the Prediction data set of 20,000 further accounts.

TIMi ranking

TIMi obtained the 7th place amongst a total of 94 teams.

PAKKD 2010 Results
Ranking	Team Name	Institution	AUC_ROC
1	GH (Grzegorz Haranczyk)	GH	0.645
2	Winners(Navin Loganathan,Ranjani Subramaniam, Shobha Prabhakar, Sankar Deivanayagam)	LatentView Analytics	0.641
2	Latentview (Priya Balakrishnan, Kiran.PV, Syluvai Anthony, Mahadevan Balakrishnan)	LatentView Analytics	0.641
4	uq (Vladimir Nikulin)	University of Queensland	0.638
5	Tabnak (Yasser Tabandeh)	Shiraz University	0.637
6	Wiggle Puppy (Daniel Felix)	Independent	0.634
7	Kranf (Frank Vanden Berghen)	TIMi	0.633
7	TZTeam (Didier Baclin) (also using TIMi)	None	0.633
7	Mjahrer (Michael Jahrer)	Commendo Consulting	0.633
10	Abhyuday (Abhyuday Desai)	Kiran Analytics,Inc.	0.629
10	iDO95 (Max Wang, Amy Yu)	Alliance Data	0.629

Source: PAKDD 2010

The lift we obtained