An adaptive density clustering algorithm based on measurement parameters and representative points

DBSCAN clustering algorithm (Density-based spatial clustering of applications with noise) is a density-based spatial clustering algorithm with noise. Due to its insensitivity to noise and its ability to cluster irregularly shaped data without specifying the number of clusters , it has been widely used. However, at the same time, DBSCAN also has some problems such as excessive dependence on input parameters, cumbersome parameter adjustment and poor clustering performance for samples with uneven density. To solve the problems in DBSCAN, the MR-DBSCANS algorithm (Adaptive density clustering algorithm based on measurement parameters and representative points) is proposed according to the newly proposed parameter adaptive dynamic selection method and representative point reallocation strategy. Firstly, constructing measurement parameters based on compactness and penalty terms. Secondly, using measurement parameters to filter noise points and core points, and selecting effective neighborhood radius(Eps), the sample density judgment threshold (MinPts), and optimal neighborhood points for the core points. Then the DBSCAN redistribution strategy based on representative points is proposed, and the clustering results will be revised through the redistribution strategy, to obtain the correct clustering results. Finally, the adaptability and clustering performance of the MR-DBSCAN algorithm are simulated and verified in the dataset