Commercial Activity Cluster Recognition with Modified DBSCAN Algorithm: A Case Study of Milan

Abstract

The clusters of stores and shops in the city are the main spatial carrier for commercial activities. For urban planners, a deep and clear understanding of the present aggregating features and the commercial activities is the fundamental premise for formulating a rational and promising planning. Nowadays, the volunteered geographic information, like POIs, provides researchers a more complete and realistic data source to analyse the commercial agglomerations. Yet, few of the researches pay attention to the scale of the commercial agglomerations while the majority of researches use density estimation method to visualize and describe the commercial agglomerations of different activity types at same scale. This paper aims to propose a modified DBSCAN method to analyse the distribution structures of commercial activity clusters through multiple scales, so as to find the optimum parameters and minPts to identify the unique aggregating features for each type of activity. The proposed DBSCAN is able to determine the global minimum points (minPts) automatically by detecting the “elbow” of the maximum cluster groups change curve through a series combination of and minPts. With the global optimum minPts, this modified DBSCAN will further find optimum from where the commercial activities form stable aggregations. In this paper, the commercial activities in Milan is taken as an example. Overall, 149234 POIs from the Milan Bureau of Industry and Commerce and Google place service are collected and be further classified into 25 categories. The result of the analysis shows that 1) commercial activities show five different types of spatial patterns: central aggregation pattern, ring around center pattern, high-density aggregation distribution, disperse distribution pattern and hierarchical distribution pattern. 2) Bars and clothing stores have the highest aggregating density of 2.7 POIs per hectare, while takeaway and repair activities have the lowest density. 3) Beauty stores and health service have the smallest unit cluster size around 3ha, the supermarkets and fuel stations have largest unit cluster size. 4) the spatial shapes of different activity agglomeration areas are varied.