Is the data normalised before model training?

Yes, all data is normalised before model training or predictions. Each input and output are scaled separately. The scaling is also applied to any data point which is used for predictions. The scaling method is pre-defined and currently there is no option for the user to change this.

Minmax scaling

For some models minmax scaling is used. Minmax scaling uses the minimum and maximum values of a column to scale it linearly to the range [0, 1]:

The following models use this method:

• Neural Network 
• K Nearest Neighbours
• Support Vector Regression

Standard scaling

For all other tabular models a standard scaler is used. Standard scaling uses the mean and the standard deviation of a column to scale it linearly:

The models using standard scaler are:

• Linear Regression
• Polynomial Regression 
• Decision Tree Regression 
• Random Forest Regression 
• Gaussian Processes Regression