A two-stage computer network packet classification method and system is proposed, which is designed for integration to a network system for classification of packets within the network system. The proposed method and system is characterized by the use of a two-stage operation for packet classification; wherein the first-stage operation involves the use of a decision-tree data module whose leaf nodes are used to store a bit vector that represents a cluster of rule groups that are located within a particular cut region in a multidimensional Euclidean space that is mapped to the field values of the input packet; and the second-stage operation involves the use of a bit-vector lookup table data module to retrieve a set of bit vectors which represent a set of possible rules in each rule group and which are intersected to find a matched rule for the input packet. This feature allows the packet classification to be implemented with low memory requirement and enhanced system performance.