Interesting Articles

Handling imbalanced datasets in machine learning

  • Imbalanced data is an issue if classes are not separable
  • For inseperable data, imbalanced data can be fixed using under/over sampling or adding extra-features
  • When there is imbalanced data it is very important took look at metrics beyond accuracy like F1 score, confusion matrix, …
  • Nice mathematical treatment of this concepts similar to what I had seen similar in estimation theory