An Improved Clustream Clustering Algorithm for Anomaly Detection in Electric Power Big Data
As one of the most important data forms, stream data has been applied to many applications, especially in electric power big data. Anomaly detection in power big data has always been an important research topic of data mining analysis. How to detect abnormal data rapidly and accurately has become a research hotspot. The poor accuracy and high complexity of the traditional detection methods, along with other limitations, make them incapable of processing modern power big data efficiently and effectively. This paper proposes an effective anomaly detection method in power big data based on the modified CluStream clustering algorithm. In the proposed method, during the online stage, Redis clusters are used to save all the data within a certain period of time and iteratively update the data over time. During the offline state, the K-means clustering algorithm is optimized to
reduce time complexity, and an optimal-distance method is used to determine the cluster centers quickly. Experiment results prove that the proposed method can accurately detect the outliers in power big data, and is quicker than the original CluStream clustering algorithm.
Keywords: Stream Data, Power Big Data, Anomaly Detection, CluStream Clustering, Online Stage, Offline Stage, K-means
Y. Wang, "An Improved Clustream Clustering Algorithm for Anomaly Detection in Electric Power Big Data", Engineering
Intelligent Systems, vol. 30 no. 3, pp. 185-193, 2022.