Analysing local spatial density of human activity with quick density clustering (QDC) algorithm
Katarzyna Kopczewska
Abstract
This paper deals with the local spatial density of human activity. By understanding and quantifying the spatial distribution of interrelated phenomena such as business location and population settlement at the micro level, it is possible to track local under- and over- spatial representation in socio-economic development. The modelling of spatial density using point data is crucial for territorially targeted policies and business decisions. Weak stream of studies in this field is a consequence of lack of methods. This study presents quick density clustering (QDC), a novel algorithm for classifying geolocated point data into low, medium and high density clusters. QDC uses two spatial features - the sum of distances to k-nearest neighbours (kNN) and the number of neighbours within a fixed radius (frNN) - to generate parameter robust, interpretable clusters. By normalising these metrics and applying K-means clustering, QDC captures both local and global density variations, making it suitable for analysing human activity at urban and regional scales. Empirical validation demonstrates its accuracy and effectiveness in partitioning point data into density clusters and comparing density groups in grids. The QDC provides a robust framework for advancing density-based studies in socio-economic research as well as environmental science and spatial statistics • Paper presents a novel Quick Density Clustering (QDC) algorithm for geolocated point density classification. • QDC combines k-nearest neighbours and fixed-radius neighbours for robust density metrics. • Method captures local and global density variations through normalised clustering. • QDC demonstrates effective density classification on simulated and empirical datasets. • Solution is applicable across socio-economic, environmental, and urban spatial analyses.