Nowadays, huge amount of data are generated and collected in many domains and from various sources. Most of the time, the collected data are processed as common data where simple calculations are applied for the analysis, such as measuring the average, the maximum, the deviation, etc. Exemplar selection has a finer meaning since its aim is to study a few exemplars from common data (the most representative ones). The objective of this paper is to propose a methodology able to extract these representative exemplars from a dataset. The proposed method has been tested against well-known simulated as well as real dataset. It is then experimented on dataset extracted from experimentations of connected vehicle traces.