Framework to deal with Imbalance Classification Problem

🔴 Imbalance Classification

🟠👉 Sampling methods are a very popular method for dealing with imbalanced data. These methods are primarily employed to address the problem with relative rarity but do not address the issue of absolute rarity.

AND

🟠👉 Sampling methods seem to be the dominant type of approach in the community as they tackle imbalanced learning in a straightforward manner .

------------------------------

🟠👉 Data OverSampling methods:

👉 Random Oversampling

👉 SMOTE

👉 ADASYN - (Adaptive Synthetic Sampling)

👉 Borderline SMOTE

👉 MWMOTE - (Majority Weighted Minority Oversampling Technique)

👉 ROSE - (Random Over Sampling Examples)

--------------------------

Will share more later, follow  Mukesh Manral🇮🇳  for more details.

--------------------------

🟠👉 Data Undersampling methods:

👉 Random UnderSampling

👉 Tomek Links

👉 NearMiss

👉 NearMiss-1

👉 NearMiss-2

👉 NearMiss-3

👉 Cluster Centroids

👉 Edited Nearest Neighbors

👉 One-Sided Selection

👉 Condensed Nearest Neighbor

👉 Neighborhood Cleaning Rule

----------------

🟠👉 Combined Oversampling and Undersampling | Hybrid methods:

👉 SMOTE and Edited Nearest Neighbors | SMOTE-ENN

👉 SMOTE and Tomek Links | SMOTETomek

👉 SMOTE and Random UnderSampling

👉 BalanceCascade

👉 G-SMOTE

👉 MC-ODG (Majority Class-Oriented Data Generation)

Refer this for more