Multi-level anomaly detection method based on exponential smoothing and integrated learning model

摘要: A multi-level anomaly detection method based on exponential smoothing, sliding window distribution statistics and an integrated learning model comprises the following steps of a statistic stage, training stage classification wherein in a, key feature set is determined according to application scene; b, for discrete characteristics, built through histogram, smoothing continuous characteristics; c, observation features all are input periodically; d, process ended. In data formed by marked normal abnormal examples; random forest trained. The provides general framework problems comprising time sequence characteristics complex behavior patterns suitable online permanent detection, used achieve advantages parallelization high generalization ability, can be applied multiple scenes like business violation telecom industry, credit card fraud financial industry network attack detection.