On Runtime and Classification Performance of the Discretize-Optimize (DISCO) Classification Approach

摘要: Using machine learning in high-speed networks for tasks such as flow classification typically requires either very resource efficient approaches, large amounts of computational resources, or specialized hardware. Here we provide a sketch the discretize-optimize (DISCO) approach which can construct an extremely classifier low dimensional problems by combining feature selection, discretization, novel bin placement, and lookup. As selection discretization parameters are crucial, appropriate combinatorial optimization is important aspect approach. A performance evaluation performed YouTube task using cellular traffic data set. The initial results show that DISCO move Pareto boundary versus runtime trade-off up to order magnitude compared optimized random forest decision tree classifiers.