Data normalization

Use the Feature Scaling activity to scale features with varying magnitudes, units, and range into normalized values.

Large differences in data ranges of numeric columns must be scaled for machine learning algorithms that use Euclidian distance when calculating the model, such as Linear Learner, or Principal Component Analysis (PCA). You must bring those features to a common scale, and simultaneously avoid range distortion or loss of data.

New values maintain the general distribution and ratios from the source data, and keep values within a scale applied across all numeric columns used in the model.