PAKDD Data Mining Competition 2009: New Ways of Using Known Methods
The PAKDD 2009 competition focuses on the problem of credit risk assessment. As required, we had to confront the problem of the robustness of the credit-scoring model against performance degradation caused by gradual market changes along a few years of business operation. We utilized the following standard models: logistic regression, KNN, SVM, GBM and decision tree. The novelty of our approach is two-fold: the integration of existing models, namely feeding the results of KNN as an input variable to the logistic regression, and re-coding categorical variables as numerical values that represent each category’s statistical impact on the target label. The best solution we obtained reached 3rd place in the competition, with an AUC score of 0.655.
Keywordsdata mining logistic regression KNN credit risk assessment
Unable to display preview. Download preview PDF.
- 1.PAKDD data mining competition 2009, Credit risk assessment on a private label credit card application (2009), http://sede.neurotech.com.br/PAKDD2009
- 3.R Development Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria (2009), http://www.R-project.org