Segmenting smart cities using biclustering algorithms

Abstract

Smart cities are characterized by the integration of information and communication technologies to improve the quality of life, sustainability, and modernity for their citizens. A key differentiator from “regular” cities is the readiness of both local authorities and residents in smart cities to adopt digital innovations and digitalization. So far, numerous cities have been characterized as “smart”; however, the question that emerges is whether all these cities are on the same level of smartness, and if not, how can they be grouped according to the level of implemented digitalization? This study investigates the segmentation of 50 smart cities with populations ranging from 600,000 to 3 million residents. Using machine learning algorithms, specifically biclustering, we analyze 14 indicators from the composite Cities of the Future index to form groups of cities with the highest degree of mutual similarity. Following this segmentation, a detailed analysis of the resulting clusters is conducted. The findings from this research can help urban policymakers develop effective smart city strategies.