Anomaly detection based on multi-dimensional kernel density estimation Epanechnikov

The present invention relates to a method of detecting abnormality based on a multidimensional Epanechnikov Kernel Density Estimation. Object is to provide a method can accurately detect abnormal data. Technical scheme: An abnormality detecting method for a multidimensional Epanechnikov nuclear density estimation, comprising the following steps in sequence: 1) All nodes each distributed data acquisition, then the value of the abnormality diagnosis using the k nearest distance based sampling methods; 2) cluster head sliding window is formed in the normal data samples, establishing kernel density estimation model in the cluster head sliding window based on the sample; 3) of the core density estimation model transmitted to each distribution node, each distribution node by using the kernel density estimation model determines whether the data arriving at the next time in each distribution node is abnormal; 4) every time T, the normal of each distribution node sends data to the latest period of the cluster head node; 5) return to step a.