Towards a R-centric architecture for multi purpose geographical analysis on heterogeneous multi-source data
Résumé
R provides an elegant and widely accepted platform with a rich function set for statistical data analysis, mathematical computations and flexible interactive graphics. The number of available R packages keeps growing at amazing speed, making it increasingly challenging for developers to deal with complex projects. A thorough study showing the interest of integration of statistical and geographical information systems has been proposed. We, then, focus on efficient computing strategies applied to general client-server applications with Geographical Information Systems (GIS) and general statistics analysis on data collected from distant or local sources. It provides an experimental platform where the complexity of the whole system involves a wide knowledge to the developper who has to deal with concurrent executions of loosely coupled tools (web servers, GIS tool, SGBD, statistics tools, ...) sharing heterogeneous data. Some types of analysis require heavy and complex computations alternating between intensive use of data in the statistical and graphical analysis under the geographical information system. This lead to the manipulation of multiple ad-hoc formats and make it necessary to define aggregated sets of heterogeneous data under R and to define multiple layers data under the GIS system. The major requirements are: ensure data coherency over loosely coupled tools, software libraries and machine-independent data formats that support the creation, access, and sharing of array-oriented scientific data, management of irregular spatial data, possibility to extend the built-in functions, database connectivity and access to local or remote file systems, integration with other software, scalability, Graphic User Interfaces (GUI), interactivity, scope of built-in data analysis functions, client/server architecture with computations being done on the server side (minimise WAN traffic) and Web interface to routine analyses.