The invention provides a data cube construction method based on data dispersion independence. The method comprises the steps that for a first tuple, 2 subsets are generated through N different attributes in the first tuple, and the 2 subsets and tuple identifiers corresponding to the 2 subsets are placed in a Hash table; for a second tuple, 2 subsets are generated through N different attributes in the second tuple; the 2 subsets generated by the second tuple are compared with existing subsets in the Hash table; if some subset generated by the second tuple is consistent with some existing subset in the Hash table, a tuple identifier corresponding to the subset is added into the list of the tuple identifier corresponding to the subset in the Hash table; otherwise, the subsets and tuple identifiers corresponding to the subsets are placed in the Hash table. The method solves the problem that the efficiency of the existing Frag-Shells algorithm is greatly lowered under a high data dispersion degree.