Standardization of features in feature engineering

Does it make sense to standardise skewed data?
So compute skewness of the set of points as listed here and if there is skewness, try taking log of the feature values and again measure skewness.
If skewness between -0.1 and +0.1 the standardise else normalise.

Hopeful of a reply.

Hello @Vijay_Rajan
I would suggest If the data exhibit skewness, it may indicate a departure from a normal distribution. In such cases, I would advise applying transformations like taking the logarithm of the feature values to help reduce the skewness and make the data more symmetric.

After transforming the data, you re-evaluate the skewness. If the skewness is within the range of -0.1 to +0.1, which indicates a distribution close to symmetry, you can choose to standardize the data.

On the other hand, if the skewness is still significant after the transformation, I can consider normalization instead.

Happy Learning
Isaak

1 Like