Clustering Multidimensional Spatial Datasets With DBSCAN OPTICS BIRCH K-Means and Two-Step Methods

<p>This book presents a systematic approach to clustering and density estimation of multidimensional real-life spatial datasets utilizing the density-based clustering methods DBSCAN and OPTICS and it compares their clustering performance to that of the traditional centroid-based K-means clustering algorithm the hierarchical BIRCH clustering algorithm and the hybrid two-step clustering algorithm (a combination of hierarchical and K-means) evaluating the quality of clustering by the five clustering approaches through five quality validation indices including the DBCV index. This book primarily provides a detailed description of the key concepts and steps involved in applying the DBSCAN and OPTICS algorithms to cluster multidimensional real-life spatial datasets interpreting the results analyzing them critically and comparing the results with those of three other popular clustering methods. The <strong>dbscan</strong> R package used for clustering with DBSCAN and OPTICS algorithms utilizes a space-partitioning data structure called a K-d tree to perform fast K-distance search and fixed-radius nearest neighbor search including all neighbors within a fixed radius thereby identifying clusters efficiently. This approach is a widely adopted robust platform for identifying arbitrary-shaped clusters in large spatial datasets. The BIRCH algorithm within the <strong>stream </strong>R package was used to efficiently cluster and identify densely populated regions in multidimensional spatial datasets delivering the best possible clustering results with minimal input/output cost. This book provides a detailed description of clustering by the five clustering algorithms used complemented by procedures for estimating and choosing input parameters inference of results and computing supported by dimension reduction techniques t-SNE using the <strong>tsne</strong> R package and principal component analysis through factor analysis in SPSS for extracting components in 2 and 3 dimensions for visual enhancement. This book will be particularly beneficial to those wishing to employ these density-based techniques in research or applications across statistics data mining and analysis clinical research social science market segmentation consumer analysis and many other disciplines.</p>