Mostly in Banking domains or credit card use cases, the data for predicting a transaction as fraudulent is extremely low due to less evidence for fraud cases resulting in an Imbalanced Dataset for ML use cases. This notebook deals with 3 techniques of handling such cases.
The 3 Techniques discussed in the notebook are :
- Under-sampling
- Over-sampling
- SMOTE Technique
Then a Random Forest algorithm is applied to check the performance of each technique.