In this work we apply low-loss electron energy loss spectroscopy (EELS) to probe structural and electronic properties of single silicon nanocrystals (NCs) embedded in three different dielectric matrices (SiO2, SiC and Si3N4). A monochromated and aberration corrected transmission electron microscope has been operated at 80 kV to avoid sample damage and to reduce the impact of radiative losses. We present a novel approach to disentangle the electronic features corresponding to pure Si-NCs from the surrounding dielectric material contribution trough an appropriated computational treatment of hyperspectral datasets. First, the different material phases have been identified by measuring the plasmon energy. Due to the overlapping of Si-NCs and dielectric matrix information, the variable shape and position of mixed plasmonic features increases the difficulty for non-linear fitting methods to identify and separate the components in the EELS signal. We have managed to solve this problem for silicon oxide and nitride systems by applying multivariate analysis methods that can factorize the hyperspectral datacubes in selected regions. By doing so, the EELS spectra are re-expressed as a function of abundance of Si-NC-like and dielectric-like factors. EELS contributions from the embedded nanoparticles as well as their dielectric surroundings are thus studied in a new light, and compared with the dielectric material and crystalline silicon from the substrate. Electronic properties such as bandgaps and plasmon shifts can be obtained by straightforward examination. Finally, we have calculated the complex dielectric functions, and the related electron effective mass and density of valence electrons.