What is the rule for deciding when to normalize Variables In pre-processing? Some techniques, Like boosting For classification, Do not require The Variables to be normalized.For other techniques, Normalization seems very important
How Do I know When I need to normalize My predictors?
Is there a general rule that I can rely on for choosing whether to scale and center? Apologies for formatting.I can only Make this Post With voice to text
 A: The most important reason to normalize your data is to bring it to the same unit (scale of [0,1]). In the end you always have to think for yourself if normalization is needed. If the method is distance based (such as KNN) or if you penalize coefficients of variables (such as in Ridge or Lasso regression), some type of scaling is needed logically.
A similar question:
When conducting multiple regression, when should you center your predictor variables & when should you standardize them? 
A: There is no golden rule for normalization. Feature scaling should be performed alongside with a model selection
Yes, in some point it is true - the tree-based models are less affected by unscaled features, as decision trees classifier tries to find the most useful split for each feature, and it won't change its behavior and its predictions.
On the other side, there are models which depend on these kind of transformations. The model based on your nearest neighbours, linear models, and neural network.
Types of preprocessing on numerical features:


*

*MinMaxScaller

*Rank

*Log transfrom

*Extracting square root

*StandartScaller


Also normalization is a good practice to remove the influence of outliers
