Abstract: Heterogeneous information network carries different protocols and network channels,and realizes the resource scheduling through the cloud storage,resulting in abnormal data which can cause security threat and storage overhead to the network information space,so accurate detection of abnormal data is essential.The traditional detection algorithms use the simplified gradient algorithm for outlier detection,which cannot effectively remove multiple abnormal data of known interference frequency components,so the detection performance is not good.An abnormal data detection algorithm was proposed based on adaptive notch cascade model.The heterogeneous information network system model is constructed,and by using the intrinsic mode decomposition,the abnormal data signal analytical model is decomposed into multiple narrowband signals.Two order lattice notch filter structure is designed,and a plurality of fixing notch filters cascade is used to suppress the interference components.Matching projection method is used to seek the optimized feature solution,and all matching feature point pairs are find out,realizing the improvement of outlier detection.Simulation results show that,when the algorithm is used in data detection,signal amplitude is larger than the amplitude of noise interference data.It can improve the detection performance,and anti-interference performance is better.

Key words: Network,Abnormal data,Detection,Adaptive notch filter