The invention discloses a classification method for an evolution rule of a cellular automaton. The method comprises the steps: outputting a code sequence based on a set cellular automaton model, carrying out the counting of cellular states of each type corresponding to to-be-classified evolution rules, calculating the equalization parameters of each evolution rule, and obtaining the type of the current evolution rule based on a type section corresponding to values of the equalization parameters. Compared with a conventional classification mode based on Langton parameters, the method enables the value sections of the equalization parameters of periodic-type, complex-type and chaotic-type evolution rules to have apparent boundaries, and effectively improves the classification accuracy of theevolution rules of the cellular automaton with the number of cellular states being greater than two.