A Pruning Neural Network Model in Credit Classification Analysis

Table 14

Extraction rate and accuracy rate comparison between PNN and other feature selection methods of the Japanese credit dataset.

Methods	Numbers of variables before feature selection	Numbers of variables after feature selection	Extraction rate (%)	Accuracy rate (%)
-test [54]	15	12	20	65.53
Stepwise [54]	15	5	66.7	82.64
Related matrix [54]	15	12	20	60.16
Factor analysis [54]	15	11	26.7	74.22
PCA [54]	15	8	46.7	74.00
PNN	15	4	73.3	85.51