Two-Stage Bagging Pruning for Reducing the Ensemble Size and Improving the Classification Performance

Algorithm 3

Distance based pruning for bagging algorithm.

Input:  -training set, - subsets sampled from , - a set of base models, - number of base models

or subsets, -feature vector representing a test sample

Output:  -a reduced set of base models, - a pruned bagging ensemble

1  Collect the subsets of out-of-bag samples as .

2  Calculate the center of each as ,

3  Calculate the Euclidean distance from the test sample to each center of

4  Given a parameter , compute the threshold , which is the td-th decile value of the set

5  Initialize

6  for    do:

7    if :

8      

9 The outcome of a test sample predicted by the pruned ensemble is given as follows: