An Automated Approach to DBSCAN Parameter Tuning for Trajectory Data Clustering

Clustering trajectories of moving objects is an important task in trajectory data analysis. DBSCAN is widely used for this purpose because it can identify clusters with arbitrary shapes and does not require the number of clusters to be specified in advance. However, its performance strongly depends on the proper selection of two parameters, eps and MinLns, which is usually done through trial and error. In this study, an automatic method is proposed to determine a suitable range for the eps parameter in trajectory clustering. First, the pairwise distance matrix between trajectories is computed and its distribution is modeled using a Gaussian Mixture Model to estimate a reasonable eps range. Then, Bayesian optimization is applied to search for the optimal combination of eps and MinLns within this range. The clustering results are evaluated using the AdapSil and the Davies-Bouldin index. Experimental results show that the proposed approach reduces manual parameter tuning and improves clustering quality.

کلیدواژه ها:

Density-Based Clustering ، Trajectory Clustering ، DBSCAN ، Automatic Parameter Selection ، Gaussian Mixture Model ، Bayesian Optimization