There are a few options for dealing with imbalanced binary classification (assuming you want to find the minority group):
-
To begin, you should examine the metrics you’ll use to assess your model. Because, and I’ll provide an example to show why, the accuracy of your model might not be the ideal statistic to look at. Let’s assume 99 of the bank withdrawals were not fraudulent, but one was. Your model would be 99 percent accurate if it just categorized every case as “not fraudulent.” As a result, you might want to think about utilizing accuracy and recall measures.
-
Increasing the penalty of misclassifying the minority class is another way to enhance imbalanced binary classification. The model should categorize the minority class more accurately by raising the penalty for doing so.
-
Finally, by oversampling the minority class or undersampling the dominant class, you can enhance the class balance.