Anomaly detection method of telecommunication traffic data based on improved dynamic clustering algorithm
Hanting Duan, Meng Hao, Xue Zhang · 2024
Aiming at the shortcomings of the existing dynamic clustering algorithm, the invention proposes an improved dynamic clustering algorithm to detect outliers of telecom traffic data. The center location of clustering is used to describe the category information, and the telecommunication traffic data is clustered to select representative data, and then the selected representative data and the original data are then clustered. In the outlier detection stage, according to the relationship between the center point and the center point of the data cluster, the position relationship between all the points in the data cluster and the center point of the cluster is calculated, and the outlier score is calculated to eliminate the abnormal data. Finally, it can seen the high efficiency of the proposed algorithm for classification tasks through experiments