Contraintes sur la prédiction des flux de neutrinos de T2K par les données de l'expérience de hadroproduction NA61/SHINE.

T2K is a long baseline neutrino oscillation experiment on accelerator based in Japan, whose primary goal is a precise measurement of the $\theta_{13}$ angle of the $P\!M\!N\!S$ matrix. This measurement is possible through the appearance of electronic neutrinos out of a muonic neutrino beam, 300~km downstream after their creation point. Neutrinos are made by the decay in flight of unstable particles (pions, kaons, muons) produced by 31~GeV/c accelerated protons impinging onto a carbon target. Most of the neutrinos produced are of muonic-type, but a non-negligible amount of electronic neutrinos is also created, which will contribute to the dominant source of errors for the measurement of $\theta_{13}$. In order to understand, and predict, this electronic contamination, a parallel hadroproduction experiment is used: NA61/SHINE at CERN reproduces the T2K beam conditions, and measures the kinematics of produced hadrons thanks to two types of target: thin and replica. The measurement of the K0S production is described in this thesis. This measurement, together with charged hadrons, is then implemented in the T2K simulation chain. The development of a simulation based on the generic tool VMC, detailed in the thesis, provides a unique framework for the simulation of the two experiments. Moreover, this tool allows tests of several hadronic models against NA61/SHINE and HARP experimental data. Fritiof-based models seem to be the most promising. All these ingredients played a key role leading to the first measurement of the $\theta_{13}$ angle, and reducing its uncertainty.