Abstract: Methane hydrates (MHs) play important roles in the fields of chemistry, energy, environmental sciences, etc. In this work, we employ the generalized energy-based fragmentation (GEBF) approach to compute the binding energies and Raman spectra of various MH clusters. For the GEBF binding energies of various MH clusters, we first evaluated the various functionals of density functional theory (DFT), and compared them with the results of explicitly correlated combined coupled-cluster singles and doubles with noniterative triples corrections CCSD(T)(F12 ^* ) method. Our results show that the two best functionals are B3PW91-D3 and B97D, with mean absolute errors of only 0.27 and 0.47 kcal/mol, respectively. Then we employed GEBF-B3PW91-D3 to obtain the structures and Raman spectra of MH clusters with mono- and double-cages. Our results show that the B3PW91-D3 functional can well reproduce the experimental C−H stretching Raman spectra of methane in MH crystals, with errors less than 3 cm ^-1 . As the size of the water cages increased, the C−H stretching Raman spectra exhibited a redshift, which is also in agreement with the experimental "loose cage - tight cage" model. In addition, the Raman spectra are only slightly affected by the neighboring environment (cages) of methane. The blueshifts of C−H stretching frequencies are no larger than 3 cm ^-1 for CH₄ from monocages to doublecages. The Raman spectra of the MH clusters could be combined with the experimental Raman spectra to investigate the structures of methane hydrates in the ocean bottom or in the interior of interstellar icy bodies. Based on the B3PW91-D3 or B97D functional and machine learning models, molecular dynamics simulations could be applied to the nucleation and growth mechanisms, and the phase transitions of methane hydrates.