Abstract : The recent starting up of the LHC calls for an update and an improvement of the theoretical predictions computed from the Standard Model. The possible signs of a new physic should at first appear as slight discrepancies between those results and the observations. As a consequence, the cross sections must be estimated anew at the LHC energy of collision then calculated at higher perturbative orders.The complexity of the manipulations involved requires the development of new, possibly numerical, techniques. Once the theoretical predictions are obtained, their precision has also to be evaluated as accurately as possible. This thesis presents developments on these three avenues. First, the already existing FONLL tool has been used to update the predictions of the heavy quark production rate at first order in perturbation theory at the energy of collision of 7 TeV. An alternative approach to the handling of a type of divergences arising in the computation of the cross section, namely the infrared divergences, is also presented. Finally, a confidence model (or bayesian model) allowing a rigorous description of the contribution of the truncation of the perturbative series to the theoretical uncertainty is detailed. A discussion on the notions of probability measure and confidence measure provides an introduction to this study.