Which technique is most effective for handling categorical variables with high cardinality to prevent dimensionality explosion?

Question

Data Science — Medium

Which technique is most effective for handling categorical variables with high cardinality to prevent dimensionality explosion?

Accepted Answer

Target Encoding replaces categorical values with the mean of the target variable for that category. This keeps the feature space dimensionality constant regardless of the number of categories. One-Hot Encoding creates a new column for every category, leading to dimensionality explosion with high cardinality. Label Encoding introduces an artificial ordinal relationship that may mislead models.