DEVI seminar- Benoît Sainct

Date - Heure / Date - Hour
Date(s) - 08/12/2016
11h15 - 12h15

Emplacement / Location
Amphi Breguet, Building Breguet

Outliers detection in tax bases.

Tax revenues, such as housing tax, represent a significant proportion of the resources of local and regional authorities in France. These incomes are mainly based on spontaneous statements by individuals about their own dwellings, which are regularly erroneous. To verify the veracity of these declarations is therefore a major issue.

The objective is to guide communities in their tax rehabilitation efforts. For this, it is necessary to establish a model of the declarations which is only based on verifiable contextual information, and then identify and measure the anomalies, i.e. buildings for which the declarative information could be erroneous.

After a few simple studies, focus will be placed on regression trees, random forests and exploiting the ordinal aspect of some variables.

Benoît Sainct, thèse Cifre, Adelyce-Institut de Mathématiques de Toulouse.