PCMIgr: a fast packet classification method based on information gain ratio

Abstract To solve the problem of ambiguous attribute selection in existing decision tree classification algorithms, a construction method based on information entropy, PCMIgr, is proposed. PCMIgr heuristic greedy strategy. At each node, when it necessary to select attributes for division, with highest gain ratio selected. The main innovation this that traditional optimized, and efficiency constructed improved compared before optimization. same time, ensures leaf node only associated one rule, which avoids common "rule replication" process construction, effectively saves memory calculation time. experimental results show application can further improve packet tree, be applied high-speed real-time classification.