[Site Map]
You are here ---
> Home
> Research
# Delta ranking : A feature selection procedure

## Delta Ranking using Java Applet

## Delta Ranking using script "R"

The delta ranking procedure is a feature selection method based on the maximization of the interclass distance between the classes, combined with a logarithmic normalization.

This version only support two classes

The Data File must contains the values separated by space, in this shape :

Case #1 - Name | Case #2 - Name | ... | Case #N - Name | |

Variable #1 - Name | Variable #1 - Value for Case #1 | Variable #1 - Value for Case #2 | ... | Variable #1 - Value for Case #N |

Variable #2 - Name | Variable #2 - Value for Case #1 | Variable #2 - Value for Case #2 | ... | Variable #2 - Value for Case #N |

... | ||||

Variable #P - Name | Variable #P - Value for Case #1 | Variable #P - Value for Case #2 | ... | Variable #P - Value for Case #N |

The Group File must contains the values separated by space, in this shape :

Group of the Case #1 - 0 or 1 |

Group of the Case #2 - 0 or 1 |

... |

Group of the Case #N - 0 or 1 |

A "R" script has been developed in order to simplify the usage of the delta ranking.

To download the script : Script R

To download the readme (How to Use) : Readme File

To download the readme example learning data file : Data File