Extraction of the essential elements for urban systems modelling – A word-to-vector approach

Abstract

Due to its ever-evolving nature, urbanisation continues to escalate in complexity, further exacerbating the urban sustainability challenges. This necessitates the need for evidence-based policymaking enabled by modelling approaches, to facilitate informed decisions, and propel and gravitate towards urban sustainability. The major constraint is that of identifying the essential characteristics for consideration when modelling cities as complex systems, in a structured manner that integrates these characteristics, cognisant of their relative importance. The distinctive urban systems, corresponding system characteristics and interdependencies impacting the modelling of cities as complex systems, can be identified from peer-reviewed literature. The limiting constraint is, although there is widely available information on cities in research databases, the ability to use this literature for a quantitative model has not been proven, presenting a research gap. This approach results in significant complexities. In order to resolve these complexities, this study seeks a systems-based approach including a 2-tier structured protocol, leveraging qualitative-to-quantitative techniques to automatically extract the key systems which impact the development of city models. Through a systematic literature review, data on 13 key systems is qualitatively extracted from research databases such as Scopus and ScienceDirect, for the duration 2014 – 2024. Through word2vector analysis, machine learning techniques are utilised to perform the quantitative mapping of each urban system into corresponding system characteristics, and quantitatively illustrate them based on relative importance. The results illustrate that this proposed method is significant to characterize the essential systems that constitute a city as a complex system, based on machine learning and text analytics.