Feature scaling normalization

• Feature Scaling (How it really works?) Explained !!
• How To Calculate the Mean and Standard Deviation — Normalizing Datasets in Pytorch
• Normalization and Standardization

Purpose of feature scaling

For most machine learning algorithms and optimization algorithms , Scaling the eigenvalues to the same interval can make the model with better performance .

for example ：

（a） There are two different characteristics , The value range of the first feature is 1_{10, The value range of the second feature is 1}10000. In gradient descent algorithm , The cost function is the least square error function , So when using gradient descent algorithm , The algorithm will obviously favor the second feature , Because its value range is larger .

（b）k Nearest neighbor algorithm , It uses European distance , It will also lead to its preference for the second feature . For decision trees and random forests as well XGboost In terms of algorithm , Feature scaling has little effect on them .

Common feature scaling methods

1. Standardization . Subtract the mean , And divide it by the standard deviation
2. normalization . Use min、max, Shrink the value to [0,1]