Normalisation/Scaling

Normalisation/Scaling#

Normalisation and scaling are the process of scaling your data to some predetermined range. It is often between 0 and 1, but can be according to a model. Normalisation and scaling are often used when comparing different types of data in routines such as cluster analysis.

The options are:

  • Interval [0 1] - This simply stretches the data values between 0 and 1.

  • Mean: zero, Standard deviation: unity - This sets the mean of the dataset to 0, and the standard deviation of the dataset to +/- 1

  • Median: zero, Median absolute deviation: unity - This sets the median of the dataset to 0, and the median absolute deviation of the dataset to +/- 1

  • 8-bit histogram equalisation [0 255] - this seeks to distribute the data equally across 255 values.

_images/rasternorm.png

Normalisation options.#