Research on Fast and Corresponding Multi-Stream Data Management System of Power Pretraining Based on Central Weighted Data Stream Clustering Algorithm

To solve the anomaly detection problem of power big data stream, this paper combines the clustering algorithm of power big data stream with power big data stream. In view of the existing stream data clustering algorithm is not easy to store all the data, power data easy to lose and other problems and the stream data clustering algorithm for offline stage clustering algorithm real-time response requirements from the data integrity, security and low time complexity of the stream data clustering algorithm to improve the CLU Stream data clustering algorithm. At the same time, a central-weighted data stream clustering algorithm is proposed. In the online phase, Redis cluster is used to buffer stream data, and node time attenuation strategy is designed. Increases the percentage of valid heartbeat messages. The offline stage clustering algorithm is optimized. The optimal distance method is used to determine the initial clustering center and reduce the number of iterations. Finally, the proposed central-weighted data stream clustering algorithm is used to detect the abnormal behavior of users. The experimental results show that the algorithm can detect the abnormal behavior of users well.