Section: Software and Platforms

Stochastic systems for knowledge discovery and simulation

The CarottAge system

Participants : Florence Le Ber, Jean-François Mari [contact person] .

Hidden Markov Models, stochastic process

The system CarottAge is based on Hidden Markov Models of second order and provides a non supervised temporal clustering algorithm for data mining. It is freely available under GPL license (see http://www.loria.fr/~jfmari/App/ ).

It provides a synthetic representation of temporal and spatial data. CarottAge is currently used by INRA researchers interested in mining the changes in territories related to the loss of biodiversity (projects ANR BiodivAgrim and ACI Ecoger) and/or water contamination. A new version incorporating a graphic user interface was released and is now running on Windows systems.

CarottAge has been used for mining hydromorphological data. Actually a comparison was performed with three other algorithms classically used for the delineation of river continuum and CarottAge proved to give very interesting results for that purpose [17] .

The ARPEnTAge system

Participants : Florence Le Ber, Jean-François Mari [contact person] .

Hidden Markov Models, stochastic process

ARPEnTAge (http://www.loria.fr/~jfmari/App/ ) (for Analyse de Régularités dans les Paysages: Environnement, Territoires, Agronomie is a software based on stochastic models (HMM2 and Markov Field) for analyzing spatio-temporal data-bases [106] . ARPEnTAge is built on top of the CarottAge system to fully take into account the spatial dimension of input sequences. It takes as input an array of discrete data in which the columns contain the annual land-uses and the rows are regularly spaced locations of the studied landscape. It performs a Time-Space clustering of a landscape based on its time dynamic Land Uses (LUS). Displaying tools and the generation of Time-dominant shape files have also been defined.

We model the spatial structure of the landscape by a Potts model with external field whose sites are LUS located in the parcels. The dynamics of these LUS are modeled by a temporal HMM2. This leads to the definition of a Potts model where the underlying mean field is approximated by a hierarchical hidden Markov model that processes a Hilbert-Peano fractal curve spanning the image.

Those stochastic models have been used to segment the landscape into patches, each of them being characterized by a temporal HMM2. The patch labels, together with the geographic coordinates, determine a clustered image of the landscape that can be coded within an ESRI shapefile. ARPEnTAge can locate in a 2-D territory time regularities and implements a Time-dominant approach in Geographic Information Systems.

ARPEnTAge is freely available (GPL license) and is currently used by INRA researchers interested in mining the changes in territories related to the loss of biodiversity (projects ANR BiodivAgrim and ACI Ecoger) and/or water contamination.

In these practical applications, CarottAge and ARPEnTAge aim at building a partition –called the hidden partition– in which the inherent noise of the data is withdrawn as much as possible. The estimation of the model parameters is performed by training algorithms based on the Expectation Maximization and Mean Field theories. The ARPEnTAge system takes into account: (i) the various shapes of the territories that are not represented by square matrices of pixels, (ii) the use of pixels of different size with composite attributes representing the agricultural pieces and their attributes, (iii) the irregular neighborhood relation between those pixels, (iv) the use of shape files to facilitate the interaction with GIS (geographical information system).

ARPEnTAge and CarottAge have been used for mining decision rules in a territory holding environmental issues. They provide a way of visualizing the impact of farmers decision rules in the landscape and revealing new extra hidden decision rules [23] .

GenExp-LandSiTes: KDD and simulation

Participants : Sébastien Da Silva, Florence Le Ber [contact person] , Jean-François Mari.

simulation, Hidden Markov Models

In the framework of the project “Impact des OGM” initiated by the French ministry of research, we have developed a software called GenExp-LandSiTes for simulating bidimensional random landscapes, and then studying the dissemination of vegetable transgenes. The GenExp-LandSiTes system is linked to the CarottAge system, and is based on computational geometry and spatial statistics. The simulated landscapes are given as input for programs such as “Mapod-Maïs” or “GeneSys-Colza” for studying the transgene diffusion. Other landscape models based on tessellation methods are under studies. The last version of GenExp allows an interaction with R and deals with several geographical data formats.

This work is now part of an INRA research network about landscape modeling, PAYOTE, that gathers several research teams of agronomists, ecologists, statisticians, and computer scientists. Sébastien da Silva is preparing his PhD thesis within this framework and is conducted both by Claire Lavigne (DR in ecology, INRA Avignon) and Florence Le Ber [46] , [40] .

GenExp-LandSiTes was part of a survey about innovative tools for geographical information [74] , [73] . This survey has been conducted within the GDR Magis and has been presented in a book both in French and in English.