Statistical Consultants Ltd

TANAGRA: A free data-mining program

Feature Type: Software Review, Data-Mining

TANAGRA is a free data-mining program. It was developed in France and released in 2004. It is the successor of SIPINA, a classification program.

TANAGRA screenshot

TANAGRA has three windows: data mining diagram, components and output. It has a ‘drag-and-drop’ type interface, where the user can drag icons (from the components window) and drop them into a nested diagram that represents a set of processes. The diagrams can be saved.

TANAGRA diagram

A right click on a component in the diagram, brings up a small menu. One of the options in that menu is ‘Execute’ which runs each component from the start of the diagram, down the hierarchy to the selected component.

**The ‘Define Status’ Component**

The ‘Define status’ component is used to define variables as being target, input or illustrative variables, for the process that follows it. For example:

In a regression model, the target variable would be the response variable, and the input variables would be the explanatory variables.
In a principal components analysis (PCA), all variables included would be input variables.
When building a classifier, the target variable is what is to be classified, and the input variables would be used to construct the classifier.

Component Categories

TANAGRA’s has the following categories of components.

Data visualisation
TANAGRA data visualisation components

Statistics
TANAGRA statistics components

Nonparametric statistics

Instance selection

Feature construction
TANAGRA feature construction components

Feature selection

Regression

Factorial analysis
TANAGRA factorial analysis components

PLS

Clustering
TANAGRA clustering components

Spv learning
TANAGRA Spv learning components

Meta-spv learning
TANAGRA Meta-spv learning components

Spv learning assessment
TANAGRA Spv learning assessment components

Scoring
TANAGRA Scoring components

Association
TANAGRA Association components

French style decimal separation

The data set is assumed to have decimal separators that conform to the French convention that uses commas instead of dots e.g. 1000,00 instead of 1000.00.

TANAGRA website:

http://eric.univ-lyon2.fr/~ricco/tanagra/en/tanagra.html