Feature Normalization

In some machine learning algorithms, failing to normalize features results in that feature being over or underrepresented

$x_{\text{normalized}}=\frac{x-\bar{x}}{s_d}$

• Essentially a z-score but using sample metrics rather than population metrics
• $x$ is the original sample data
• $\bar{x}$ is the mean of $x$
• $s_d$ is the standard error in $x$
• This is the equivalent of the StandardScaler() in sklearn

$x_{\text{normalized}}=\frac{x-x_{\text{max}}}{x_{\text{max}}-x_{\text{min}}}$