Discretization and moments

Some data analysis algorithms require categorical input data instead of numeric input data. In this case, the data must be preprocessed through a discretization step in which numeric values are mapped to discrete values. Moments are quantities that describe some aspects of continuous attribute distributions. Of particular interest are central moments or moments around the mean.

Discretization

Discretization algorithms are divided in the following categories:

Unsupervised: The target concept or class attribute is not used for setting the interval bounds.; Typically, unsupervised algorithms are applied when discretization is done without the intention of later using the data for classification, or when different classification tasks can be considered for the same data set. Here, different attributes are used as the target concept.
Supervised: The target concept or class attribute is used for setting the most appropriate interval bounds.; Conceptually and computationally more complex than unsupervised algorithms.

Moments

Moments belong to the task family of data exploration.

The purpose of data exploration is:

Becoming familiar with the data
Detecting possible problems regarding the quality of the data
Observing data distributions that might be useful for subsequent analytical processing

Although data exploration is usually followed by predictive modeling, it might produce useful results on its own.