An Improved DBSCAN Algorithm for Adaptively Determining Parameters in Multi-density Environment
Feiya Chen · 2021
The density-based spatial clustering DBSCAN algorithm can realize clustering of arbitrary shapes in noisy data sets. The DBSCAN algorithm needs to manually determine the Eps and Min Pts parameters. As a result, the accuracy of the clustering results directly depends on the user's choice of parameters. In the face of large-scale irregular data, fixed parameters lead to ineffective multi-density data clustering. ideal. In response to the above situation, an improved DBSCAN algorithm for adaptively determining parameters in a multi-density environment is proposed. The non-parametric kernel density estimation theory is used to analyze the distribution characteristics of data samples to automatically determine the Eps parameters, preprocess the data, and identify each Based on the density around the data object, the density threshold MinPts suitable for the density of the area is automatically generated. The manual intervention of the clustering process is avoided, and the automation of the clustering process is realized. This method can select reasonable Eps and MinPts parameters, and obtains clustering results with higher accuracy.