Cosmic rays are highly energetic subatomic particles coming from outer spaces. Among them one may find electrons, which may carry valuable information on how were they produced and accelerated. To study them among other objectives, the DAMPE satellite was sent into Earth orbit. It has been taking data nonstop since then and has already revealed new structures in the spectra of cosmic rays. Such a complex detectors warrants an equally powerful analysis method : for this thesis, I developed an electron identification method based on neural networks. Reliable, the classifier retains high efficiency at all energies and significantly outperforms traditional methods. The classifier is integrated in an analysis chain to fully understand DAMPE data. Our uncertainties stay below 8.5 % as we detect up to 14 TeV electrons. Their spectra exhibits a hardening at 3.6 TeV, potential hint of a local source of cosmic rays.