Due to the large scale of uncertain data and limited accuracy in clustering, the efficiency of data clustering is low. Therefore, a density peak based uncertain data clustering algorithm is proposed. Using the Mahalanobis distance method, interfering sample data is eliminated with low correlation, the missing values of uncertain data samples is calculated through entropy, and gradually the reverse restoration is performed. Using the density peak calculation method the distribution of cluster centers is determined. A decision graph is introduced to partition data clusters, the K-nearest neighbor idea is used to calculate the trust values of non cluster center data samples, secondary identification and partitioning of data points and noise with large trust value differences within clusters, optimizing the density peak clustering method. The experimental results show that when facing large-scale data, accurate clustering can still be achieved with less clustering time. The proposed method has high computational efficiency and has great significance for uncertain data mining and analysis.

Markov distance method, Entropy value, decision diagram, peak density, K-nearest neighbor thought