The most stable and accurate Categorical encoders are target-based encoders. Target encoding is the process of replacing a categorical value with the mean of the target variable. From the Source 4. Regularization is a must for target-based encoders.
Several Kaggle Competitors use mean encoding and Target encoding with regularization to predict much better and rise through ranks in the leaderboard.
Why is regularization used in target encoders? Is regularization a must?
