Data Preparation - Standardization／Standardizing／Normalizing Data Variables ／ Min-Max／Mean／Z-Score Feature Normalization／Scaling

given 𝒙 = {𝑥₁, …, 𝑥_𝑛}

where:

• 𝜇 - mean
• 𝜎 - standard deviation

Z-Score Normalization Advantages

• standardized variables are centered at 0 so that we don’t need to compute the y-axis intercept for linear regression
• in gradient descent learning algorithms learns the model coefficients “equally” in multiple linear regression
• the slope is then exactly the same as the correlation coefficient, which saves another computational step