Bioinformatic Random Seed

Keep up with what's happening in Bioinformatics and Machine Learning (^ω^)

Back to main page

Class Imbalance

In a binary classification senerio, if 1) cost of false positive and false negative is the same, 2) negative class is predominant, the classifier would always predict negative.

Strategy for solving class imbalance problem

  1. Cost function based approach
    • Synthetic Minority Oversampling Technique (SMOTE)
    • Adaptive Synthetic (ADASYN)
  2. Sampling based approach
    2.1 Upsampling
    2.2 Downsampling