## Section: New Results

### Data Mining for Material Flow Analysis: Application in the Territorial Breakdown of French Regions

One of the major issues for assessment of the long-term sustainability of urban areas is related to the concept of “imported sustainability”. In order to produce such an assessment for a given territory, one must first identify and quantify the types of materials used, and the impacts associated to these uses. Material Flow Analysis (MFA) is directly related to how the material circulates and how it is transformed within a territory. In most cases this analysis is performed at national and regional levels, where the statistical data is available. The challenge is to establish such an analysis at smaller scales, e.g. in the case of France, at the department or city level.

We have explored the possibility of applying data analysis at the regional level by generating a mathematical model that can fit well the data at regional scale and estimate well the departmental one. The downscaling procedure relies on the assumption that the obtained model at level `n' (for example region) will be also true at level `n+1' (for example department), such that it could properly estimate the unknown data based on a set of chosen drivers (socio-economic data). We have designed and implemented techniques based on parameter optimization and model selection as well as robust estimation, in order to estimate the best drivers for a given set of territories, i.e. the socio-economic data (e.g. employees per type of manufacturing industry, population data, etc.) that best correlate with the production of various types of agricultural or other products [19] .