The fast searching clustering algorithm of the density peak is a simple and efficient density-based clustering algorithm. However, there are shortcomings such as the setting of the truncation distance c d is too sensitive, the similarity measure is too simple, and the artificial selection of the cluster center points is subjective. To deal with these problems, this paper proposes a new density peak clustering algorithm KE-DPC (KNN-ESD-density-peak-cluster) that can automatically select the cluster center points. First, the algorithm uses the near information to adjust the distribution of data samples, and optimizes the similarity measurement criterion in combination with Euclidean distance. Then the local density calculation formula is redefined according to the number of neighbor samples, thereby avoiding the setting of the sensitive c d . Finally, the sample distribution on the decision map is fitted by linear regression to obtain the Residual set, and the cluster center point is automatically obtained according to the characteristics of the Residual analysis in ESD anomaly detection, removing the subjectivity of artificial selection. The experimental results of the artificial data set and UCI standard set show that the KE-DPC algorithm is better than K-means, DBSCAN, DPC, A-DPC and other algorithms.