An efficient approach for network traffic classification

Classifiers fail to handle high network traffic and changing node behaviors in efficient manner. Such applications require incremental learning algorithms with low computational complexity and low misclassification rate. This paper presents a model which partitions the training set into equivalence classes on the values of each feature. During classification, algorithm picks up one partial solution set per feature from respective equivalence partition. It collates weak classifiers, thus obtained to classify the test instance in partially lazy manner. The algorithm scores over contemporary classifiers in terms of better complexity for classification, incremental learning complexity and low misclassification rate.