Handling unbalanced data¶

What is unbalanced data?¶

• Large discreptancy between "positive" and "negatrive" cases
  • i.e., fraud detection. Fraid is rate and most rows will be non-fraud.
• Mainly a problem with neural networks.

Oversampling¶

• Duplicate samples from the minority class
• Can be done at random

Undersampling¶

• Instead of creating more positive samples, remove the negative ones
• Throwing data away is usually not the right answer
  • Unless you are specifically trying to avoid "big data" scaling issues.

SMOTE¶

• Synthetic minority over-sampling technique
• Artificially generate new samples of the minority class using nearest neighbors
  • Run KNN of each sample of the minority class
  • Create a new sample from the KNN result (mean of the neighbors)
• Both generates new samples and undersamples majority class
• Generally better than just oversampling

Adjusting thresholds¶

• When making predictions about a classification (fraud / not fraud), you have some sort of threshold of probability at which point you'll flag something as the positive case (fraud).
• If you have too many false positives, one way to fix that is to simply increase that threshold.
  • Guaranteed to reduce false positives
  • But, could result in more false negatives