Résumé
This paper presents a research on the estimation of the impact of geometric imprecision on basic measurements (length, area) in vector databases, in order to generate relevant information for decision making. The goal consists in the elaboration of a model allowing a non-expert user to evaluate the geometric imprecision of its dataset, using data analysis as well as description of production processes (such as digitising or generalisation). We suppose that these processes induce variable contribution to errors in a dataset, and are exposed to spatial heterogeneity according to the geographical context. This model lays on a knowledge base, based on measurements, contextual indicators and additional information on the dataset production. In order to evaluate a dataset's geometric imprecision impact without any reference, decision rules are under development, using the knowledge base coupled with hypothesis on the influence of the geographical context on production processes. Experimentation on a road network illustrates the respective impact of production processes in the final length measurement error. Possibilities to communicate this impact following a particular usage are also evocated.