Environemental data mining

Preferred relationships with the House of Remote Sensing (MTD) of Montpellier have allowed us to be confronted with new data which represent challenges for data mining community: satellite data, land, presence of experts, local skills. Thus, by proposing new approaches for symbolic clustering, we have, in the context of food security in Mali , offered not only prediction rate significantly higher than the traditional approaches (75% vs 60%) but also the ability to automatically create hierarchies to summarize the data . These techniques are promising to handle large volumes of data such as those provided in the context of Équipex Géosud whose objective is the provision of a high resolution annual nationwide satellite coverage. We are particularly interested in the detection of changes for these heterogeneous spatio-temporal data. The ANR Fresqueau project whose aim is to monitor the quality of rivers has allowed us to propose new approaches to model large amounts of data but also the implementation of new specific patterns (Thesis Michael Fabregue) .

Last update on 01/12/2014