Combination of Density-Based Spatial Clustering With Grid Search Using Nash Equilibrium

Abstract

This paper introduces a novel clustering approach that enhances the traditional Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm by integrating a grid search method and Nash Equilibrium principles and addresses the limitations of DBSCAN parameterization, particularly its inefficiency with big data. The use of Nash equilibrium allows the identification of clusters with different densities and the determination of DBSCAN parameters and the selection of cells from the network, and significantly improves the efficiency and accuracy of the clustering process. The proposed method divides data into grid cells, applies DBSCAN to each cell, and then merges smaller clusters, capitalizing on dynamic parameter calculation and reduced computational complexity. The performance of the proposed method was assessed over 3 big-size and 11 middle-size datasets. The achieved results implied the superiority of the proposed method to DBSCAN, ST-DBSCAN, P-DBSCAN, GCBD, and CAGS methods in terms of clustering accuracy (purity) and processing time.