Continuous variables segmentation and reordering for optimal performance on binary classification tasks

It is common to And continuous input variables with non-monotonic propensity relation with the binary target variable. In other words, when taken as propensity scores, these variables do not generate unimodal Kolmogorov-Smirnov Curves. However, these variables possess highly discriminant information which could be explored by simple classifiers if properly preprocessed. This paper proposes a new method for transforming such variables by detecting local optima on the KS2 curve, segmenting and reordering them to produce a unimodal KS2 on the transformed variable. The algorithm was tested on 4 selected continuous variables from the benchmark problem of Loan Default Prediction Competition and the results showed significant improvement in performance measured by both the AUC_ROC and Max_KS2 metrics for 3 different Artificial Intelligence algorithms, namely Linear Discriminant Analysis, Logistic Regression and MultiLayer Perceptron.