Network flow self-learning method based on multi-engine comprehensive analysis

The invention discloses a network flow self-learning method based on multi-engine comprehensive analysis. The method comprises the following steps: S1, collecting network flow; s2, preprocessing the network traffic data; s3, forwarding the network traffic data preprocessed in the step S2 to a network traffic self-learning multi-engine integrated framework, and carrying out deep and multi-dimensional mining analysis learning; wherein the network flow self-learning multi-engine integration framework comprises a learning detection engine based on a feature rule, a learning detection engine based on deep learning and other network flow learning detection engines which can be flexibly integrated; s4, network traffic anomaly mining: detecting and mining abnormal data in the network traffic according to a network traffic anomaly detection rule in a learning detection engine based on a feature rule; and S5, network flow anomaly analysis and alarm: carrying out analysis, research and judgment on the detected network f