[Retracted] The Use of Hellinger Distance Undersampling Model to Improve the Classification of Disease Class in Imbalanced Medical Datasets

Algorithm 1: Hellinger Distance Undersampling (HDUS) pseudocode

Input: Imbalanced Training dataset (ITrD)

Output: Balanced Training dataset (BTrD)

1 Group the ITrD according to the classes

2  C1= ITrD (class1) //C1 indicates the minor class which contains less number of instances

3  C2= ITrD (class2) //C2 indicates the major class which contains more number of instances

4  For i in rows of (C2)

5   For j in rows of (C1)

6    Sim_i,j = calculate the similarity between C2(i) and C1(j) using Hellinger Distance

7     append Sim_i,j To HD(i)

8   Next j

9    select m top values from HD (i) // where m is a given number of neighbouring minority class

10    HD_sum(i)= sum the selected m top values

11   Next i

12 C2HD=select w majority class instances according to the highest similarity value in HD_sum(i),

  // where w is a given number

13 return (BTrD= C2HD +C1)

Algorithm 1: Hellinger Distance Undersampling (HDUS) pseudocode