TIMi top winner at the PAKDD 2007

TIMi top winner at the PAKDD 2007 The PAKDD2007 is the 11th Pacific-Asia Knowledge Discovery and Data Mining conference. These results were published the 1st may 2007. Problem description: Cross-Selling: Credit Card 2 Mortgage. We try to sell a “mortage” to customers that already have a “credit card”. Difficulty: Target is small: 1.71% Data: 40,700

TIMi top winner at the PAKDD 2007

TIMi top winner at the PAKDD 2007

The PAKDD2007 is the 11th Pacific-Asia Knowledge Discovery and Data Mining conference.

These results were published the 1st may 2007.

Problem description:

  • Cross-Selling: Credit Card 2 Mortgage.
  • We try to sell a “mortage” to customers that already have a “credit card”.

Difficulty:

  • Target is small: 1.71%

Data:

  • 40,700 Card customers with the company within a specific 2-year
  • 700 Targets (1.71%): opened a home loan with the company within 12 months after opening the credit card
  • 40 modelling variables

Evaluation:

  • Prediction dataset without target (8000 customers)
  • Criterion: Area under the lift curve (AUC)

TIMi Model:

  • 17 variables.
  • Quality on build (AUC) ~= 73.26 % ± 1.8

 

The lift of our model

TIMi obtained the 6th RANK of the PAKDD2007 competition amongst 47 competitors.

Get in touch


Stop renting compute power. Start owning your company.

Discuss your constraints, your volumes, your governance requirements.

We will show you how TIMi fits into your architecture.

Opens in a new tab