Traditional packet classification algorithms based on space-decomposition usually use only one heuristic to split the rule space, and they don't adopt different heuristics according to the characteristics of each dimension. This paper proposes a hybrid intelligent cutting (HIC) scheme for packet classification. HIC firstly partitions the ruleset according to the IP prefix length. Then, taking into account the characteristics of current cutting dimension in each subruleset, HIC uses bit cuttings and precise projection point cuttings to cut the IP dimension and port dimension, respectively. At last, HIC builds the decision tree of hybrid cutting structures. Simulation results show that HIC has better scalability with different rulesets. Compared with EffiCuts, its time and space performance have increased by 46% and 74% respectively.