Instance selection

Concepts

These components reduce the set of actives examples before an analysis. Examples is subdivided in active and inactive examples.

**If the learning process is realized on active examples, projection (new attribute)
is always applied to all the examples.**

Attributes status

None.Instance selection components

Component | Description | Parameters | Note |

Sampling |
Sampling.
Sequential sampling, see Vitter's work. |
- Size of sample: proportion or number of examples. | |

Stratified Sampling |
Stratified sampling |
- Stratification attribute, a discrete one. - Stratification type : balanced or proportional. - Size of the sample: proportion or number of examples. |
A discrete attribute must be available in the dataset. |

Recover examples |
Permutation of the status of examples: actives become inactive, and conversely. | - Set to be active: inactive or all examples. |

Last modification : January 21st, 2004.