Feature transformation for Exponential distribution I am a beginner in machine learning, and working on a classification task. I am spending lots of time doing feature engineering, and I understand that for most models one would like to scale the features so that they are relatively Normal and have values close to 0 (or at least I believe that).
This is fine for features which are approximately Normal, however I have some features which follow an exponential distribution. How should I proceed to engineer these features? Does the engineering change depending on what model I use?
 A: 
relatively Normal and have values close to 0 (or at least I believe that).

Not true.  For one thing, most methods don't care about the distribution of the predictors.  This confusion likely comes from assumptions about the distribution of the outcome.  Second of all, this would be impossible for binary predictors.

How should I proceed to engineer these features? Does the engineering change depending on what model I use?

Yes and no.  Some methods benefit from normalizing (subtracting the sample mean and dividing by the sample standard deviation).  Others can handle data on the original scale just fine.  My own approach is to always normalize continuous variables.  If the variable has a very large scale and there are several outliers, it may be worth taking the log or the square root of the variable and then normalizing.
A: Well, I depends on the type of Machine Learning model you will use. If this model is linear, such as Logistic Regression or LDA, it is better to engineer this variable before training the model. If the model is a Decision Tree, Random Forest for example, the initial distribution doesn't affect much the outcome as the splitting technique is optimized by the own model.
In the case of a linear model, I suggest you to apply a Yeo-Johnson transformation on this variable. You can find the details in Wikipedia and it is implemented and easy to use in Scipy.
