Network intrusion detection method based on multi-empirical kernel learning

The invention discloses a network intrusion detection method based on multi-empirical kernel learning. The method comprises the following steps: learning a preprocessed unbalanced network intrusion detection sample to obtain a midpoint sample of positive and negative samples with universal gravitation balance and a neighbor sample of the midpoint sample; combining middle point samples and neighborsamples of the universal gravitation balanced positive and negative samples with multi-empirical kernel learning; and generating two regularization items in each kernel space, using generated regularization items corresponding to midpoint samples of the universal gravitation balanced positive and negative class samples for fitting classification boundaries, and using neighbor samples of the universal gravitation balanced positive and negative class samples for disturbing and correcting formation of the classification boundaries; and finally, performing voting on a classification result of each kernel space to obtain