Light Gradient Boosting Machine-Based Link Quality Prediction for Wireless Sensor Networks

Algorithm 2

Borderline-SMOTE algorithm for data imbalance.

Input:(Training sample set), (majority class label), (minority class label).

Output: New training sample set S.

1:  for every in do

2:    if the label of in

3:      Put into majority class sample set

4:    else

5:      Put into minority class sample set

6:  for every in do

7:    Calculate its nearest neighbors from the whole training set;

8:    Calculate the number of samples belonging to the majority class in neighbor samples, denoted as ;

9:    if

10:      Put into boundary sample set ;

11:    else

12:      Put into safe sample set ;

13:  for every in do

14:    Synthesize the new minority samples by using (2);

15:  Merge the synthesized minority sample with the original training

sample to generate a new training sample set S;

16:  return: new training sample set S.