Research on abnormal data mining algorithm based on ICA

The research of abnormal data mining is very important for ensuring the reliable operation of data mining. The algorithm currently used in data mining of anomaly mining is only a kind of measuring error time or spatial correlation. Considering the shortcomings of the existing outlier data mining algorithms and the temporal and spatial correlation of the error matrix, an abnormal data mining algorithm based on ICA is proposed. In this algorithm, the ability of reducing the dimension of BSA and the multi-scale modeling ability of wavelet transform are used to construct the algorithm of abnormal data mining. In the analysis of residual anomaly data, it is realized mainly by EWMA and Shewart control chart. And the sliding window mechanism is applied to realize the online expansion of outlier data mining algorithm, and online ICA outlier data mining algorithm is obtained. By analyzing the outlier data and the simulation results, we can conclude that compared with the BSA algorithm and the KLE algorithm, the ICA algorithm has more outstanding advantages and better detection performance.