Name
Abstract | ||
|---|---|---|
Packet classification is a very basic technique in supporting various advanced network services, for example, network measurement, quality of service, flow routing, and so on. Traditional decision tree-based packet classification algorithms often generate many redundant rules. To solve this issue, we propose GroupCuts, which employs two novel ideas. (1) Space relationship-based rule clustering: according to the analysis of classifiers, GroupCuts clusters the rules of similar size and then partitions the classifiers into subgroups. Building decision trees in these sub-groups achieves fine space performance. (2) Dynamic Point Split: in order to reduce the rule duplication in decision trees, we select multiple rule projection points to accomplish space decomposition. Using simulations of 1,000 to 100,000 rule classifiers, we show that, compared with other existing algorithms, the proposed algorithm achieves an improvement in memory requirement without reducing the search performance. |
Year | DOI | Venue |
|---|---|---|
2013 | 10.1109/GLOCOM.2013.6831260 | GLOBECOM |
Keywords | Field | DocType |
clustered dynamic point split algorithm,network measurement,space relationship-based rule clustering,pattern clustering,quality-of-service,quality of service,packet classification,decision tree-based packet classification algorithms,next generation internet equipment,rule duplication reduction,internet,multiple rule projection points,advanced network services,memory optimization,flow routing,telecommunication network routing,groupcuts,decision tree,decision trees,dynamic point split,space decomposition,clustering algorithms,classification algorithms,algorithm design and analysis | Decision tree,Data mining,Cluster (physics),Algorithm design,Computer science,Quality of service,Algorithm,Flow routing,Cluster analysis,Statistical classification,Packet classification | Conference |
ISSN | Citations | PageRank |
2334-0983 | 1 | 0.35 |
References | Authors | |
8 | 2 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Weitao Han | 1 | 1 | 0.69 |
| Yi Peng | 2 | 1303 | 78.20 |