Use SMOTE and ROSE algorithms to balance data

Use both SMOTE (Synthetic Minority Over-sampling Technique) and ROSE (Random Over-Sampling Examples) algorithms to balance data. SMOTE is implemented within KNIME. ROSE can be accessed via R.

It is advisable to balace only your training data and leave the test/validation data as they are or you run the risk of greatly inflated values on your precision statistics.

External resources

• more options about unbalanced data
• SMOTE Hints from KNIME Team members 3 (Classification Threshold Analysis)
• SMOTE Hints from KNIME Team members 2
• SMOTE Hints from KNIME Team members 1
• Imbalanced Data : How to handle Imbalanced Classification Problems
• forum entry
• R ROSE