The most common task in data analysis is to find relationships of a set
of input parameters to one or more output parameters. This kind of
analysis is called supervised because the goal of the analysis is given by
the user. In the case that no explicit goal parameter is available, the
analysis is called unsupervised. The machine learning framework
package offers a wide variety of methods for supervised and
unsupervised analysis of data.
The following examples describe several methods and scenarios in
which supervised or unsupervised analysis can help to understand the
underlying structure of your data. These methods can also be used to
create computational models from your data sources.
Demonstrates the general tasks of supervised analysis, including how to
create a decision tree and a set of rules to describe your data and use
these rules to forecast the classification of new data.
Describes several methods available for unsupervised learning, including
how self-organizing maps (SOMs) can be computed, how different
clustering methods can be used to find regions of similarity in the
data, and how supervised learning methods are applied to create
descriptions for the clusters found.
Shows how a one- and a two-dimensional function can be approximated
using various supervised learning methods, including SOM, decision tree,
rule based, and RENO.
Note: If you do not own a copy of Mathematica, you will need to download Wolfram CDF Player in order to
view these examples.