Both Normalization and Standardization are used for Feature Scaling. Below are the differences between them.
Standardization |
Normalization |
| Data should follow a Normal Distribution. | It is not required for data to follow a Normal Distribution. |
| The value lies from -x and +x. There is no boundary range. The mean and standard deviation of standardized data is 0 and 1 respectively. | The value lies between 0 and 1 or -1 and +1. |
| Data can be standardized using methods like z-score. | Data can be normalized using methods like MinMax Scaler. |
| Mean and Standard Deviation are used for Scaling. | Minimum and Maximum values are used for Normalization. |
| Standardization is slightly affected by outliers. | Normalization is slightly highly affected by outliers. |
Z – Score Formula
A z-score tells you where the value lies on a normal distribution curve.
z-score =
where
μ = mean
σ = standard deviation
x = current value
MinMaxScaler Formula
MinMaxScaler is a normalization method that uses the maximum and minimum values of a distribution to calculate the normalized value.
where
x = current value
xmin = minimum value of the distribution
xmax = maximum value of the distribution