An efficient DBSCAN optimized by arithmetic optimization algorithm with opposition-based learning
Yang Yang,Chen Qian,Haomiao Li,Yuchao Gao,Jinran Wu,Chan-Juan Liu,Shangrui Zhao
DOI: https://doi.org/10.1007/s11227-022-04634-w
2022-11-12
Abstract:As unsupervised learning algorithm, clustering algorithm is widely used in data processing field. Density-based spatial clustering of applications with noise algorithm (DBSCAN), as a common unsupervised learning algorithm, can achieve clusters via finding high-density areas separated by low-density areas based on cluster density. Different from other clustering methods, DBSCAN can work well for any shape clusters in the spatial database and can effectively cluster exceptional data. However, in the employment of DBSCAN, the parameters, EPS and MinPts, need to be preset for different clustering object, which greatly influences the performance of the DBSCAN. To achieve automatic optimization of parameters and improve the performance of DBSCAN, we proposed an improved DBSCAN optimized by arithmetic optimization algorithm (AOA) with opposition-based learning (OBL) named OBLAOA-DBSCAN. In details, the reverse search capability of OBL is added to AOA for obtaining proper parameters for DBSCAN, to achieve adaptive parameter optimization. In addition, our proposed OBLAOA optimizer is compared with standard AOA and several latest meta heuristic algorithms based on 8 benchmark functions from CEC2021, which validates the exploration improvement of OBL. To validate the clustering performance of the OBLAOA-DBSCAN, 5 classical clustering methods with 10 real datasets are chosen as the compare models according to the computational cost and accuracy. Based on the experimental results, we can obtain two conclusions: (1) the proposed OBLAOA-DBSCAN can provide highly accurately clusters more efficiently; and (2) the OBLAOA can significantly improve the exploration ability, which can provide better optimal parameters.
computer science, theory & methods,engineering, electrical & electronic, hardware & architecture