BPNN anomaly data detection algorithm based on gray wolf algorithm to optimize K-means clustering

Aiming at the situation that the K-means clustering algorithm tends to fall into the local optimal solution during the clustering process, and the clustering results are prone to errors, this paper proposes a clustering algorithm based on gray wolf optimization-means, Realize the initial selection of K-means cluster centers through the global optimization ability of the gray wolf optimization algorithm. And update the cluster centers through iterative wolf $\alpha$ to optimize the K-means clustering algorithm. Aiming at BP neural network as a supervised learning algorithm, prior knowledge of data is required for training, due to the different data types generated by different events, the applicability of BP neural network is not strong, the paper proposes a combination of K-means clustering algorithm based on gray wolf algorithm optimization and BP neural network. Cluster the initial data set through K-means clustering, and label the clustered data, then import the labeled data as a training set into the BP neural network for training, and obtain the final detection model to realize online detection of large amounts of data. The experimental results show that the algorithm proposed in this paper on the IBLK dataset and the Taihu Lake water quality dataset is compared with the traditional K-means clustering algorithm and the random forest algorithm based on firefly optimization proposed in [12] in the IBLK dataset and Taihu Lake. Experimental verification on the water quality data set, the detection rate was increased by 8.9%, 17.7% and 1.15%, 12.6%; the false alarm rate was reduced by 8.1 %, 19.3% and 1.12%, 13.6% respectively.