Feature Engineering Made Easy

Z-score standardization

The most common of the normalization techniques, z-score standardization, utilizes a very simple statistical idea of a z-score. The output of a z-score normalization are features that are re-scaled to have a mean of zero and a standard deviation of one. By doing this, by re-scaling our features to have a uniform mean and variance (square of standard deviation), then we allow models such as KNN to learn optimally and not skew towards larger scaled features. The formula is simple: for every column, we replace the cells with the following value:

z = (x - μ) / σ

Where:

• z is our new value (z-score)
• x is the previous value of the cell
• μ is the mean of the column
• σ is the standard deviation of the columns

Let's see an example ...