Urban resilience and livability performance of European smart cities: A novel machine learning approach

Abstract

Smart cities are centres of economic opulence and hope for standardized living. Understanding the shades of urban resilience and livability in smart city models is of paramount importance. This study presents a novel two-stage data-driven framework combining a multivariate metric-distance analysis with machine learning (ML) techniques for resilience and livability assessment of smart cities. A longitudinal dataset for 35 top-ranked European smart cities from 2015 till 2020 applied as the case study under the proposed framework. Initially, a metric distance-based weighting approach is used to weight the indicators and quantify the scores across each aspect under city resilience and urban livability. The key aspects under city resilience include social, economic, infrastructure and built environment and, institutional resilience, while under urban livability, the aspects include accessibility, community well-being, and economic vibrancy. Fuzzy c-means clustering as an unsupervised machine learning technique is used to sort smart cities based on the degree of performance. In addition, an intelligent approach is presented for the prediction of the degree of livability, resilience, and aggregate performance of smart cities based on various supervised ML techniques. Classification models such as Naïve Bayes, k-nearest neighbors (kNN), support vector machine (SVM), Classification and Regression Tree (CART) and, ensemble models including Random Forest (RF) and Gradient Boosting machine (GBM) were used. Three coefficients (accuracy, Cohen's Kappa (κ) and average area under the precision-recall curve (AUC-PR)) along with confusion matrix were used to appraise the performance of the classifier ML models. The results revealed GBM as the best classification and predictive model for the resilience, livability, and aggregate performance assessment. The study also revealed Copenhagen, Geneva, Stockholm, Munich, Helsinki, Vienna, London, Oslo, Zurich, and Amsterdam as the smart cities that co-create resilience and livability in their development model with superior performance.

Introduction

With an estimated population growth of 6.7 billion in cities globally by 2050, multifaceted intelligent urban systems form the norm (Sun et al., 2020, Sun et al., 2020a). In the midst of unfettered urban flux, Laissez-faire urbanization has drenched the leapfrogging possibilities of smart solutions and digital intelligent platforms to turn cities into more liveable units thus, failing to offer a dignified standard of living to the urban inhabitants (Calzada, 2017). Digital solutions provide opportunities for development, at the same time pave ways for abuse and entail a litany of challenges (DeRolph et al., 2019). The use of smart technologies in cities have intensified beyond borders of utilitarianism to the extent of implying pressure on infrastructure (Lee et al., 2021). When rethinking strategic autonomy in the digital era, smarter cities, a paradigm beyond smart cities framed to optimize challenges, present a mesmeric case in tremendously ameliorating interconnectivity with less focus on creating value for urban inhabitants (Boykova et al., 2016). These smarter cities are moving up the ladder of digital development where techno-centricity takes the driver's seat (Yigitcanlar and Lee, 2014). Digital solutions in cities scale with users (Hatuka and Zur, 2020). However, embarking on technological development beyond a point where technology has met the user requirement, involves risk. Despite smart technologies being a prerequisite in intuitively bridging gaps and concerns of urban inhabitants, the bureaucratic barriers have led to an uncoordinated drive for the technologies to scale when attempting to engage city residents for these technologies to work (Ramboll, 2020). Smart city experience of today focus on city dwellers as a means for testing smart solutions with less concern being paid on their values, beliefs, and livability (Mouratidis, 2021). When scaling technological developments within the urban context, the concept of livability requires special attention, as people and their interactions are the key drivers for technologies to find their application in the smart ecosystem (Sutriadi and Noviansyah, 2021). To continue, without human settlements, cities don't exist and thus focusing on the concerns of city residents and including the dimensions of people and communities with economic and environment pillars when addressing the concept of smart city is crucial for sustainable outcomes. However, being smart and sustainable too cannot fully improve all the key “quality-of-life” aspects and foster livability (Yigitcanlar and Lee, 2014).

The concept of livability can transform intelligent units into habitable spaces (Pan et al., 2021). However, techno-centric development must not only focus on livability as the soul to an endurable unit, but also on the ability of a city to rebound post stress, thus offering a dignified standard of living to the urban inhabitants. Cities are often vulnerable to unexpected predicaments such as economic upheaval, anthropogenic disruptions, climate change, geopolitical instability, public health crisis, and diplomatic embargos (Ukkusuri et al., 2021). Smart cities of today despite realizing the importance of resilience are no exception to these uncertainties. The Covid-19 outbreak in Wuhan, central China is a classic exhibition of insufficient city resilience (Chu et al., 2021). The Covid-19 pandemic has left lime lighted questions on urban resilience and livability of tech-driven smart cities around the globe (Feng et al., 2022). The pandemic paradigm has left opportunities for smart, sustainable, and mega-cities to optimize urban systems to cope with future external disruptions for a sustainable, liveable, and resilient habitable unit. Thus, it is seen that the failure of holistic thinking in optimizing development challenges has resulted in sceptics raising concerns on whether the aspiring smarter cities of today or the cities that claim to be smart and sustainable hold a striking balance between livability and resilience. Thus, the central agenda when making incremental technological improvements in smart and sustainable cities must include the traditional involvement of people and strategies to address the challenges in an urban scale; in short, resilience and livability. Thus, co-creating livability and resilience in cities to scale artificial intelligence (AI) solutions in a sustainable manner has become a top priority. It is incredibly important to explore and understand how smart cities address the concept of resilience and livability and to what extent ranked smart cities address these paradigms in planning for next generation cities. It is unclear so as to, ‘Where does the world cities that are smart, smarter, and smart-sustainable fall within the dimensions of resilience and livability?’

Smart cities are complex urban ecological systems built to optimize challenges and improve the residents quality of life with the ubiquitous use of data (Alsarayreh et al., 2020; Shehab et al., 2021; Kutty et al., 2022). However, the vision of smart cities goes beyond the use of information and communication technologies (ICT) for better resource use and fewer emissions for most (see: Albino et al., 2015; Mohanty et al., 2016; Akande et al., 2019; Sharifi, 2019). While, it means smarter urban transport networks, upgraded water supply and waste disposal facilities and more efficient ways to light and heat buildings for some (see: Ahvenniemi et al., 2017; IEEE, 2022). It also means a more interactive and responsive city administration, safer public spaces, capacity to absorb, recover and prepare for future shocks and, meeting the needs of an aging population for others (see: UNESC, 2016; ITU, 2016). Despite the literature explaining the concept of smart cities abundantly, with the concept expanding over the years, currently there is still not yet a common and acceptable definition for the smart city concept (Arafah, and Winarso, 2017). To support the implementation of the multifarious smart city concept, it is crucial to measure the performance of cities with the aim to historically document their strengths and weaknesses, for the scope of future improvements and to inform interested stakeholders about the level achieved in different target goals. Smart city assessment tools and models present city-rankings, revealing the best (and the worst) places for certain activities, which is pointed out by literature to be a central instrument for assessing the attractiveness of urban regions. Several smart city assessment tools and frameworks exist such as the smart ranking systems developed by the University of Vienna, the Intelligent Community Forum's Smart21 communities, the Global Power City Index, the Smarter Cities Ranking, the World's Smartest Cities, the IBM Smart City, and the McKinsey Global Institute rankings (Albino et al., 2015). However, an exhaustive overview of the frameworks, rating systems, and number of indicators, for the aforementioned smart city assessment tools and models, conducted by Albino et al. (2015), revealed a lack of thematic focus on the multi-dimensional concept of the expanding smart city concept and the omission of the indicator typologies. 43 indicator frameworks were scanned for indicators that could be related to the CITYkeys pre-selected subthemes and thus potentially be used for the CITYkeys framework. Based on this inventory analyzed by Neumann et al. (2015), it is reported that in general terms, the analyzed frameworks suggested that the availability of the key performance indicators (KPIs) were saturated. It also reported the following gaps in terms of the indicator availability: multilevel governance and economic vibrancy, education, employment, scalability, accessibility, and replicability. The report suggests that there was a significant variation in the coverage of different sub-themes, including for instance the “energy and mitigation” and “environment” sub-themes (Neumann et al., 2015). To continue, there have also been several approaches to standardize the indicators from which the frameworks or rankings can provide an assessment for smart city implementation. Recently, Huovila et al. (2019) provided a quite extensive comparative analysis of existing standardized indicators for smart city assessment. The analysis provided by Huovila et al. (2019) indicated that there is a lack of balance between the different indicators, namely between the indicators related to sustainability and smartness. There is also a strong emphasis on the smartness indicators (ISO-37120:2018, 2018; Sharifi, 2019). Huovila et al. (2019) affirms that International Organization for Standardization (ISO), European Telecommunications Standards Institute (ETSI), and Sustainable Development Goal 11 are well documented, but the International Trade Union (ITU) standards have a short definition of the indicators. The well-known indicator sets proposed by the ITU under the ‘United 4 Sustainable Smart City’ (U4SSC) initiative to shape future cities focus only on integrating sustainability with smartness under the dimensions: Economy, Environmental and, Society and Culture (ITU, 2016). Despite targeting the soul agenda of urban smartness and sustainability, the initiative promises on making future cities more resilient. However, the indicators under U4SSC fails to address urban resilience in depth across multiple dimensions of resilience.

Thus, it is seen that many of the existing smart city assessment frameworks and tools are mainly used for promotional purposes and very few for an evaluation of what actually should be done in order to increase the performance of future developments in terms of resilience and livability along with urban smartness. When a city is planned to be smart, it is a must to prepare the city to be resilient at all times (Arafah and Winarso, 2017). Further, to mitigate the negative effects of urbanization in cities, it is an essential target of sustainable smart cities to transform dwelling units into livable spaces (i.e., urban livability) - a concept clearly underrepresented in the smart city frameworks analyzed (Benita et al., 2021). Thus, envisioned to discretely answer the call of improved local livability and susceptance to unexpected predicaments, it is seen that the effective implementation of city resilience and urban livability face numerous obstacles. A recent review published by Ramirez Lopez and Grijalba Castro (2021) on resilience in smart cities revealed a lack of integrity in practically addressing the multi-dimensional facets of city resilience, where a biocentric vision of territorial urban planning and capacity building is undertaken than a human centric approach to better living. Similarly, a review conducted by Paul and Sen (2020) revealed that the developed western cities that often act as benchmarks for Dickenson cities account the livability from a physical aspect (such as mobility options, transit-oriented design, and fiscal supremacies) than from a socio-economic perspective. It is to note that, shocks are meant to occur within cities that are termed smart, but to what speed can the cities rebound to their natural state is a question that institutions and policy makers must answer to better protect cities when under chaos. This requires a standardized lens for city leaders to analyse the resilience capacity to position adaptation to unexpected predicaments and livability frame of reference to envision a human centric development targeted for better living standards.

In vain, the livability and resilience paradigm have been used interchangeably in several contexts targeting the soul agenda; quality of life with a smart growth strategy to rebound post-stress. Given the intrinsic element of kinship between urban resilience and livability, it is crucial for planners and policy makers to analyse these paradigms under a generalized frame of reference tailored across multiple aspects. Thus, investigating smart city development from a broader strategic vision in light of resilience capacity and livability is crucial. For the same, several assessment approaches exist such as the non-parametric optimization based techniques like the data envelopment analysis, composite index based scoring, GIS and remote sensing based assessments and many more. Machine learning (ML), a subset of artificial intelligence (AI) has recently gained immense attention owing to its ability to effectively determine the relationship between the input features and the response variable (s) in a complex system (Wakjira et al. 2021, 2022; Hwang et al., 2021). Despite their great capability, machine learning models are rarely applied in the field of resilience and livability assessments in an urban scale. In this research, a large dataset of smart cities has been collected and used to propose a novel machine learning based framework for the assessment of resilience and livability of smart cities. Several machine learning classification and predictive models are built to understand the level of resilience and livability of smart cities based on a pre-defined set of indicators under multiple aspects. Thus, the ML algorithms would predict whether a city/smart city with specific ranking is resilient, livable and whether or not they co-create resilience and livability in their urban development model. To this end, this research targets to achieve the following objectives as to;

• Present a novel two-stage framework combining metric-distance based multivariate analysis with machine learning techniques for the assessment of urban livability and resilience of smart cities based on various influential indicators.

• Conduct a comprehensive assessment of city resilience and livability of 35 leading European smart cities as the case to identify their coping capacities based on their clustered performance as high, medium, and low.

• Predict the degree of livability and resilience, as categorical variables, based on the values of the indicators under each aspect of resilience and livability using machine learning classifiers.

• Compare and select the best classifiers based on coefficients such as model accuracy and precision to predict the degree of aggregate performance as the classification output.