Balance the training data

aksaha
aksaha New Altair Community Member
edited November 2024 in Community Q&A
What is the best way to balance data for the predictive model using deep learning algorithm when the positive class in the training dataset is just 25% (not balanced)  ?
I used 10 fold cross-validation to create the training and validation dataset.
Tagged:

Best Answer

  • varunm1
    varunm1 New Altair Community Member
    Answer ✓
    Hello @aksaha

    You can use the "SMOTE" operator present in the operator toolbox extension of rapidminer. This helps you balance the classes by upsampling the class with low samples. Use this smote operator in the training side of cross-validation.

    In my view, 75:25 is not highly imbalanced as it is close to real-world scenarios, but if you want to upsample the data you can use SMOTE as mentioned above. 

Answers

  • varunm1
    varunm1 New Altair Community Member
    Answer ✓
    Hello @aksaha

    You can use the "SMOTE" operator present in the operator toolbox extension of rapidminer. This helps you balance the classes by upsampling the class with low samples. Use this smote operator in the training side of cross-validation.

    In my view, 75:25 is not highly imbalanced as it is close to real-world scenarios, but if you want to upsample the data you can use SMOTE as mentioned above.