In materials science there is an increasing need for developing a robust and reliable first-principle approach capable of predicting crystal structures, by taking only the stoichiometry as an input. We integrate several methodologies to tackle this problem including quantum chemistry cluster calculations, simulated annealing algorithm for structure modelling, density functional theory total energy calculations and symmetry group analysis. A case study is Mg(BH4)(2) in the aim to find the reasons for discrepancies between theoretically and experimentally proposed structures. In addition to new stable monoclinic, orthorhombic and tetragonal structures, a cubic one is suggested as a possible high energy structure. Moreover, the symmetry group analysis makes possible to link symmetry-related structures via group-subgroup relations, and subsequently identify local minima on the Potential Energy Surface.