Grid-Based Density Clustering with Adaptive Parameter Tuning for High-Dimensional Data

Abstract

Grid-Based Density Clustering Algorithm is an efficient and scalable approach in data mining that integrates the strengths of grid-based and density-based clustering methods. Unlike traditional clustering techniques that operate directly on individual data points, this method partitions the data space into a finite number of non-overlapping cells, forming a grid structure on which clustering is performed. By calculating the density of each cell, the algorithm identifies dense regions representing potential clusters while filtering out sparse areas as noise. This significantly reduces computational complexity and enhances performance, making it suitable for handling massive, high-dimensional, and noisy datasets. Well-known algorithms such as STING, CLIQUE, and WaveCluster demonstrate the effectiveness of this approach in discovering clusters of arbitrary shape and size. Widely applied in spatial data mining, image analysis, bioinformatics, and market research, grid-based density clustering continues to evolve as a robust tool for large-scale data analysis and knowledge discovery.