Built LDA and QDA models on variables obtained from Principal Component Analysis (PCA) and Kolmogorov-Smirnov (KS) and tuned by leave-one-out cross-validation (LOOCV) to predict fraudulent online advertising click traffic