Research On Density Peak Clustering Algorithm Based On Natural Nearest Neighbor

Aiming at the problem that the density peak clustering algorithm is sensitive to parameters,the noise points cannot be eliminated in advance,and the clustering results obtained by processing complex manifold data are poor,this paper proposes a density peak clustering algorithm(GD-SNR-DPC)that integrates geodesic distance and shared natural neighbors.Firstly,the idea of natural nearest neighbor is used to remove noise points.Secondly,the local density of the data object is described by the shared mutual natural neighbor similarity based on the shared natural inverse neighbor set,and the initial clustering center of the data set is selected by the fast sorting algorithm,and the natural nearest neighbor is used to redefine the geodesic distance between the initial clustering centers.Then,the local density calculated by the shared mutual natural neighbor similarity and the geodesic distance newly defined by the natural nearest neighbor are used to select the initial clustering center on the decision graph,and the final clustering center is selected by the decision graph.Finally,the remaining data objects are assigned to the cluster where the cluster center is closer than its large distance.Research analysis and experiments show that the new algorithm proposed in this paper has better advantages and higher accuracy for complex manifold data clustering.Aiming at the problem that the existing uncertain clustering algorithms also have high parameter sensitivity and low clustering effectiveness for complex manifold uncertain data,this paper proposes an uncertain data density peak clustering algorithm based on JS divergence(UDPC-JS).Firstly,the uncertain natural neighborhood density factor defined by uncertain natural nearest neighbor is used to remove noise points.Secondly,the local density of uncertain data objects is described by the combination of uncertain natural nearest neighbor and JS divergence.Then,the initial clustering centers of uncertain data sets are found by combining the concept of representative points,and the distance between the JS divergence and the graph of each initial clustering center is redefined.Then,the final clustering center is selected on the decision graph by using the newly defined JS divergence and graph distance between the initial clustering centers based on the local density calculated by the uncertain natural nearest neighbor and JS divergence.Finally,the unassigned uncertain data objects are assigned to the final cluster center according to certain rules.Through theoretical analysis and research,it is proved that compared with the experimental comparison algorithm,the proposed algorithm can solve the parameter sensitivity problem well and has more significant advantages in clustering uncertain data sets of complex manifolds.