Misdeclaration self-adapting network safety situation predication method

The invention relates to a misdeclaration self-adapting network safety situation predication method, comprising the following steps: (1) extracting alarm events in a safety protection software; (2) eliminating misdeclaration in the alarm events based on a system host and network abnormal information to form an exact training sample set; (3) training the sample set by using a neutral network learning algorithm to build a predication model; (4) performing on-line predication and confirming the predication result; (5) if the predication result is misdeclaration, marking the current predication event sequence to be negative example, implementing increment neutral network learning and adjusting the predication model. By utilizing the method, the problems that too much many misdeclaration exist in the network safety situation predication and cannot be eliminated automatically are solved, the network safety situation predication model training sample set is built exactly, the predication model is built effectively, the predication result is confirmed automatically to eliminate the misdeclaration and adjust the predication model automatically, the number of misdeclaration generated in subsequent predication is reduced, and the reliability and practicability of the method are enhanced.