Table 4 Prediction accuracy: w/o categorical variables, w/ categorical variables, PCA and Binning
From: An alternative approach to dimension reduction for pareto distributed data: a case study
Experiment | Dimension of the categorical space | AUC-ROC Ten-fold cross validation | Standard deviation | AUC-ROC testing | # of Meter Devices | |
|---|---|---|---|---|---|---|
defective | Non-defective | |||||
#1 | No categorical variable | 85% | 1.7% | 86% | 2062 | 15,652 |
#2 | 205 | 78% | 8.5% | 83% | 2062 | 15,652 |
#3 | 128 | 73% | 12,5% | 81% | 2062 | 15,652 |
#4 | 48 | 76% | 13.4% | 85% | 2062 | 15,652 |