Notebook might not open because of heavy visuals, you can see it on kaggle also https://www.kaggle.com/shweta2407/oversampling-vs-undersampling-techniques

How to classify a highly unbalanced or skewed data ?

An unbalanced data or skewed data is the dataset that has its most of the data falling in one class and rest in others.

To classify this type of data, we need to first balance the data.

How to balance the unbalanced data ?

Apply different resampling techniques to balance the data : there are 2 kinds of resampling techniques - OVERSAMPLING & UNDERSAMPLING techniques.

OVERSAMPLING Techniques

SMOTE - Synthetic Minority Oversampling Technique

UNDERSAMPLING Techniques

NearMiss Version 1, 2, 3

Tomek Links

Condensed Nearest Neighbor

Edited Nearest Neighbor

Combination of Oversampling & Undersampling Techniques

One Sided Selection (Tomek Links and the Condensed Nearest Neighbor (CNN)

Neighborhood Cleaning Rule (Condensed Nearest Neighbor & Edited Nearest Neighbors )