Please enable JavaScript.
Coggle requires JavaScript to display documents.
Encoders – A Short Summary - Coggle Diagram
Encoders – A Short Summary
Classic and Contrast encoders
These do not use the information from the dependent variable in the encoding.
If there are k unique values in a categorical variable, they create up to k distinct columns to store the encoding depending on the technique.
Ordinal
One-Hot
Binary
Hashing
Helmert
Backward difference
Polynomial
Bayesian encoders
These use information from the dependent variable in the encoding. One must ensure that the encoding is only done after
splitting the data into training and test sets
and using the target from the training set. These encoders always create one column to store the encoded value. It is often recommended to add some ‘noise’ to the resulting encoded value.
Target
Leave one out
Weight of evidence
James Stein
M Estimator
Types Of encoding
Nominal
One Hot Encoding
One Hot Encoding with many variables
Mean Encoding
Ordinal
Label encoding
Target Guided Ordinal Encoding