1 Introduction

In an era marked by rapid urbanization, city planners and authorities across the globe find themselves facing an increasingly complex challenge: to not only accommodate the swelling populations of their metropolises but to also enhance the quality of life for their residents. Local governance authorities use various strategies to attract more residents, as population growth can contribute to a city's economic development [1, 2]. The pursuit of this goal necessitates a deep understanding of the urban quality of life (UQOL) attributes that contribute to a city's desirability, a quality that not only attracts more inhabitants but also fosters a sense of satisfaction and contentment among citizens.

The main research questions in our study include: what are the key attributes that affect the perceived desirability of cities, and how do these attributes interact to shape the UQOL, making certain cities more attractive to residents than others? In this context, this paper aims to unravel the attributes contributing to perceived desirability of cities to live. It uses association rule mining (ARM) methodology to shed light on the characteristics that reinforce the attraction of large cities. The use of ARM is not merely confined to the identification of individual factors in isolation; rather, it investigates the connected relationships between various elements that collectively shape the citizens’ desire for living in a city. By utilizing the power of ARM algorithm, we embark on a novel approach that allows us to discover intricate patterns and associations among these attributes, revealing hidden synergies and dependencies that might otherwise go unnoticed. This holistic methodology offers a comprehensive view of how different facets of urban life interact and combine to create an environment that residents find appealing.

Moreover, our study goes beyond finding the interrelationships of the existing factors contributing to urban desirability. We introduce the innovative concepts of order qualifiers and order winners, concepts borrowed from the realm of operations management. These concepts lend a fresh perspective to the understanding of urban development, highlighting the critical attributes that are essential for a city to be considered competitive and desirable, as well as those that distinguish a city as a standout choice among its peers. By integrating these concepts, our research offers a multifaceted understanding of the dynamic forces that drive the success of urban environments, empowering urban planners and decision-makers with insights that can influence the trajectories of their cities.

By conducting a thorough examination that combines survey-based data reflecting residents' perspectives and statistical records from the largest cities in the United States, our study provides a comprehensive portrayal of the determinants influencing urban desirability. We investigate the intricate relationship between economic, social, and environmental attributes in this regard. Additionally, we investigate the pivotal role of possessing a nationwide top quality as a potential competitive advantage, examining whether it qualifies as an 'order winner' capable of favoring one city over another in terms of desirability.

The findings of our research carry important implications for urban governance, policymaking, and the development of future cities. An in-depth comprehension of the interrelated elements fostering urban desirability empowers urban planners to strategically design environments that enhance residents' desire for city living.

The rest of the paper is organized as follows: Sect. 2 reviews the existing literature. Sections 3 and 4 describe the dataset used in the analysis and details of the methodological approach applied in this paper, respectively. Section 5 discusses the results and findings. Section 6 concludes the paper and suggests directions for future research considering the current model limitations.

2 Literature review

Understanding the key attributes that are influential in making a city a more desirable place to live is critical for urban governance authorities to make effective development decisions. Despite the difficulty, understanding the UQOL attributes has received extensive attentions in the literature [3,4,5,6]. Economic factors like median salary and household income are also investigated to have impacts on the perceived quality of life by citizens [7]. Other economic factors are considered to have impacts on life satisfaction. Diener and Suh [8] identified the cost of living as a key indicator of quality of life, finding a significant correlation between life satisfaction and living expenses. Furthermore, Motlhatlhedi et al. found that a high long-term unemployment rate negatively affects well-being [9]. Social factors are prominently featured in the literature. Zhao and Tang emphasize the importance of urban safety and security in enhancing the quality of life and fostering sustainable development in urban areas [10]. Nadai et al. [11] investigate the interplay between crime, socio-economic factors, the built environment, and mobility characteristics of neighborhoods across various cities. Zin et al. [12] examine the influence of various social factors, such as education and residential community, on quality of life. Health-related aspects are also underscored in this context [13, 14]. Additionally, commute time emerges as a critical element in assessing urban quality of life, especially in large cities, where it reflects congestion levels and its potential impact on a city's attractiveness as a place to live [15, 16]. Environmental factors such as climate is also investigated as an important factor that could affect the satisfaction of residents to live in a city [17, 18]. In this paper, we examine key UQOL attributes such as unemployment rate, median salary, crime rate, cost of living, household income, education, healthcare, climate, commute time, and community wellbeing, as identified in the literature. Additionally, we explore the impact of offering top-quality national amenities in a city on its attractiveness to residents, which is discussed in more detail in Sect. 2.3. The existing literature in this field lacks a comprehensive study that uncovers the interrelationships among UQOL attributes. Predominantly, current studies have concentrated on examining linear relationships. Literature review in this paper is structured around three main streams as follows:

2.1 Subjective vs. quantitative UQOL attributes

A major group of studies used citizens’ perceptions to detect the relationships between subjective attributes of quality of life in metro areas [19,20,21,22,23,24]. These studies relied on public opinion by using survey-based datasets. Zhang et al. investigate the impacts of urban residents’ perceptions on public rental housing as a dimension of residents’ quality of life and well-being [25]. Aminian et al. use survey methods to assess urban quality of life. They measure urban satisfaction focusing on the built environment through behavioral indicators [26]. In the same category, some used behavioral decision theory methods such as reviewing literature, interviewing experts, interviewing residents, etc. [6, 27].

Another cluster of studies focused on the analysis of urban living conditions using quantitative data at the city level [28,29,30,31,32,33]. In this category, researchers primarily relied on census data and information from governmental departments concerning various aspects of cities. Brunelle utilizes Statistics Canada Census data spanning from 1971 to 2006. It focuses on the employed population to analyze industry and regional effects in the changing spatial division of functions across Canadian urban areas. The data type used is primarily quantitative, involving detailed census data that allows for a comprehensive analysis of industry and occupational trends over time within urban areas [28]. Burnley indicates the use of quantitative data, primarily drawn from the 1986 Census of Population and Housing. This data is utilized to analyze outmigration trends from metropolitan New South Wales to non-metropolitan areas, focusing on economic, environmental, and lifestyle factors [29].

Some studies also examined combining both survey and secondary databases [6, 34]. It is suggested in the literature that both citizens’ perceptions about living conditions as well as city statistical records are required to achieve a more effective analysis of quality of life [30, 35, 36]. McCrea et al. suggest that further investigation is needed to explore the connections between objective and subjective indicators of UQOL [37]. Wirth et al. also encourage studies that employ models to connect objective indicators of the urban environment with the subjective assessments of its inhabitants [38]. The current study has used both types of datasets in assessing quality of life attributes in U.S. metro areas to not only directly incorporate public feelings into the assessment, but also allow constructive comparisons between different local areas based on municipal official records. In addition, this paper includes a city’s top-quality characteristics to investigate its association with desirability of the city to live in.

2.2 Individual vs. combined level UQOL

Among studies investigated the important factors that affected the desirability of cities, the majority has studied the quality-of-life factors in an individual level [39,40,41,42]. The common methodology applied in these studies is statistical modeling such as multiple regression and logistic regression analysis. The individual-level analysis evaluates each attribute independently; however, it lacks an investigation of how the criteria are associated with each other. Single level measures are not capable of fully capturing interrelations between UQOL dimensions [43].

Recognizing the linkage between key factors will help decision makers to understand what the real priorities are [44]. However, there are limited studies that addressed the associations between UQOL key attributes together [30, 45]. The limitation of the existing studies in proposing the combined measurement is that they only focused on the evaluation of a specific city from different dimensions [46, 47], while the comparison of the cities, nationwide, has not been considered. To fill this research gap, the current paper proposes the application of a data mining methodology that finds interesting sets of association rules. Each rule illustrates a set of UQOL attributes that together will lead to higher desirability of cities. It explains that the UQOL measurement should be analyzed in a combined level of factors, rather than an individual-level evaluation.

2.3 Urban desirability qualifiers and winners

The competition between cities to obtain a nationwide competitive advantage was highlighted in different areas in the literature. Some examples of such areas are tourism industry [48], small business sector [49], and sustainability [50]. However, no study has analyzed the competitive advantage in the context of quality of life, to investigate its impacts on residents’ desires. This paper investigates key characteristics that qualify a city to be desirable. In addition, it examines the effects of providing a top-quality characteristic as an indicator of competitive advantage, on desirability. The closest to this study, is the literature that explored the linkage between city branding and attractiveness of places to live, on a global level [51, 52].

In summary, there is a notable gap in the literature regarding a model that can effectively uncover the hidden interdependencies among UQOL attributes, which collectively influence citizens' satisfaction. This paper introduces a novel approach by applying ARM in the context of UQOL. It is pioneering in emphasizing the significance of establishing minimum standards, referred to as urban desirability qualifiers, for cities to be considered highly desirable. Furthermore, this study introduces an innovative variable, termed the top-quality characteristic, to investigate whether excelling nationally in certain attributes enhances a city's attractiveness. This approach represents a significant advancement in understanding the complex dynamics that contribute to urban desirability and quality of life.

3 Data

The data used in this study includes the 99 largest cities in the United States–cities with a population greater than 200,000–in 2016. To employ a multidimensional framework and examining a wide range of characteristics that are associated to the desirability of cities, the data was collected from various sources including U.S. national government agencies, municipal databases, and open public surveys.

Details regarding the sources of each data set are provided in Sect. 3.1.

3.1 Variables

The assessment of the composite indicator of city desirability index involves the consideration of many UQOL factors simultaneously. The extensive scientific literature from diverse disciplines such as business, politics, healthcare, environmental engineering, sociology, geography, psychology, etc., testifies to the importance and complexity of this matter. In this study, we examine the three core aspects of the city desirability index, namely economic, social, and environmental indicators [53], and we also added the dimension of top-quality characteristics to examine its association with the desirability of cities.

We collected data from various sources to employ a multidimensional framework and examine a wide range of UQOL characteristics. The data used in this study includes the 99 largest cities in the United States, with a population greater than 200,000, in 2016. We collected data from various sources, including U.S. national government agencies, municipal databases, and open public surveys provided by the U.S. News & World Report. By using a combination of survey and secondary databases, we not only directly incorporate public perceptions into the assessment, but also allow constructive comparisons between different metro areas based on quantitative city-level data.

The following variables are used to measure the economic dimension of the quality of life in a city:

3.1.1 Unemployment rate

This variable is used to measure the relationship between the labor market situation and the desirability of a city for residents. The unemployment rate is studied as an important economic attribute in a variety of research [17, 23, 54,55,56]. A growing unemployment rate is found to decrease the overall community happiness level. The data used in this study is based on the U.S. Bureau of Labor Statistics' annual unemployment rate data [57] in 2016 for measuring the local labor market situation.

3.1.2 Median salary

The relationship between salary and overall desirability of cities has been thoroughly examined in numerous studies [17, 31, 43, 58]. These studies have consistently found that higher income levels are associated with a better quality of life. This paper utilizes data from the 2016 U.S. Bureau of Labor Statistics’ Employment and Wage Estimates (OEWS) program [59]. The OEWS dataset covers almost 800 occupations, providing a comprehensive view of income across various job types.

3.1.3 Median household income

Median household income is a key factor in the study of overall UQOL. The relationship between residents' affluence and the desirability of living in a city is examined in studies such as Li and Weng [31]. Household income, as defined by the U.S. Census Bureau, encompasses the total income of all occupants in a residential unit who are 15 years or older, regardless of their relationship.

To analyze affluence in this paper, we utilize data from the American Community Survey (ACS) released by the United States Census Bureau in 2016 [60]. This dataset provides valuable information on household income, allowing for a comprehensive analysis of residents' economic status and its impact on overall quality of life.

3.1.4 Cost of living

The cost of living is a crucial factor in assessing the quality of living conditions in a metropolitan area [43, 61,62,63]. In this study, the local cost of housing services is used as a measure of the cost of living. To determine the cost of living in various U.S. cities, data from the ACS published by the U.S. Census Bureau in 2016 is utilized [64]. This dataset provides valuable information on the cost of housing services, allowing for a comprehensive analysis of the cost of living and its impact on overall quality of life in different metropolitan areas.

To assess the social aspects of quality of life, various community variables are considered, including the availability of quality health care, wellbeing, crime rate, and access to quality education.

3.1.5 Availability of quality health care

The availability of quality health care is an important consideration in the literature on quality of life, as noted extensively in studies such as [65,66,67], Access to quality health care is examined to measure the relationship between accessibility of quality health services in an area and perceived quality of life.

To measure availability of quality health care, we use data from the 2016 U.S. News Best Hospitals rankings is used [68]. This data is calculated based on the quantity of ranked medical care facilities within 50, 100, and 250 miles of each metropolitan area, as reported by the U.S. News and World Report. By utilizing this data, we can assess the availability and accessibility of quality health care services in different metropolitan areas and their potential impact on overall quality of life.

3.1.6 Community wellbeing

The community wellbeing factor is a crucial consideration in measuring the social aspect of quality of life in urban areas [17, 20, 69]. It assesses the extent to which residents have access to opportunities for good quality local amenities.

To measure community wellbeing score for each city, we use data published in 2016 by ShareCare [70]. This data captures citizens’ perceptions of community wellbeing. The index includes physical and mental health, convenient accessibility to health, food, fitness, etc. By utilizing this data, we can measure the extent to which residents have access to opportunities for wellbeing in different metropolitan areas.

3.1.7 Crime rate

Crime rate is an important factor in evaluating perceived quality of life and residents’ desire to live in a certain area. It is extensively studied in the literature as a key indicator of safety in a community [23, 71,72,73]. Ensuring the safety of citizens from the threat of violence is a fundamental human right. This paper uses data from the 2016 U.S. Federal Bureau of Investigation’s Crime Data on violent events [74] to measure the level of safety in urban areas.

3.1.8 Access to quality education

Access to quality education is an important variable in evaluating the desirability of a city, as perceived by its residents [17, 75]. This study examines whether the provision of better education is associated with a higher level of city desirability. To evaluate access to quality education, the study uses data from the 2016 U.S. News Best High Schools dataset [76], which provides rankings for schools in different geographic areas. The dataset is a reliable measure of the quality of education available to residents in each metro area.

The environmental dimension of urban living was evaluated in this study by examining two variables, namely commute time and climate.

3.1.9 Commute time

Commute time is a critical factor in assessing the quality of life, particularly in large cities, as it measures the level of congestion and its potential impact on the desirability of a city as a place to reside [15, 16, 63, 71, 77]. To investigate this factor, data from the American Community Survey published in 2016 by the U.S. Census Bureau was utilized [78].

3.1.10 Climate

Climate is an important factor that contributes to the desirability of cities [17, 18, 79, 80]. To assess climate comfort, we use the Camelot Climate Index (CCI) published by comparative climatic data for the cities of United States in 2016 [81], is used to measure climate comfort, which is a composite index that takes into account nine different elements including maximum and minimum temperatures, average annual snowfall and rainfall, average percent of sunshine, average number of days with precipitation, average number of days with minimum temperatures below 32 ºF and maximum temperatures above 90 ºF, and average relative humidity, to determine an overall score.

It is postulated that people's preference for a particular city can be influenced by the comparison of its qualities with those of other cities. To investigate this, we included the top-quality dimension in our analysis.

3.1.11 Top quality

We sought to determine whether offering superior qualities than all competitors could result in cities being more desirable to their residents, as perceived by them. We introduced a binary variable to indicate whether a city is top-quality in a characteristic or not. This study represents the first attempt, to our knowledge, to explore top-quality characteristics in a municipal context.

3.1.11.1 Exploring the correlation among UQOL variables

To assess the possibility of redundancy in the UQOL variables, we computed the correlation coefficients between them. Correlation coefficients among variables are calculated to measure the strength and direction of the linear relationship between each two variables in the dataset. The Pearson’s correlation method is used as follows:

$$r=\frac{\sum \left({x}_{i}-\overline{x }\right)\left({y}_{i}-\overline{y }\right)}{\sqrt{\sum {\left({x}_{i}-\overline{x }\right)}^{2}\Sigma {\left({y}_{i}-\overline{y }\right)}^{2}}}$$
(1)

where:

  • \({x}_{i}\) and \({y}_{i}\) are the individual sample points indexed with i.

  • \(\overline{x }\) and \(\overline{y }\) are the means of the x and y variables.

  • The summations run over all data points.

Table 1 displays the correlation coefficients among the UQOL variables, which fall within the range of ± 0.5. There is no value above 0.5 or below − 0.5, hence there are not strong correlations among variables. There are moderate correlations among some variables. The correlation among costs of living with median salary (0.408), household income (0.405), commute time (0.496) are some examples of this category. Health Care, Education, and Top-quality show weaker correlations with other variables.

Table 1 Correlation Coefficients Among UQOL Variables
Full size table

3.2 Measuring the degree of desirability of cities

To assess the level of desirability associated with various cities, we utilized data obtained from a survey conducted by U.S. News & World Report in 2016. The survey, which involved more than 3000 participants across the United States, directly queried respondents about their desire to reside in specific cities. Notably, U.S. News and World Report is ranked as the fourth most frequently referenced and read report by U.S. journalists [82].

3.3 Data preparation

The data preparation process involved several steps. First, the UQOL variables were categorized as above-average or below-average. Cities with below-average values for cost of living, unemployment rate, crime rate, and commute time were considered to perform better than average in these areas, while cities with above-average values for median salary, median household income, availability of quality education, accessibility of quality health care, community wellbeing, and climate were considered to perform better than average in these areas.

The top-quality characteristic attribute was represented by a binary variable, with a value of 1 indicating that a city was top-rated in a particular UQOL factor, and a value of 0 indicating that it was not. For example, San Diego, CA received a value of 1 for the climate attribute because it has the greatest number of days with pleasant weather compared to other U.S. metro areas.

To measure the desirability level of living in a city, a score between 0 and 10 was used. A score of 0 indicated the lowest desirability, while a score of 10 indicated the highest desirability. Highly desirable cities were defined as those with scores in the top one-fifth slot (scores ≥ 8), based on the 5-point Likert scale used to categorize participants' evaluations into five categories. Values in the top one-fifth group were considered to represent highly perceived qualities.

4 Methodology

4.1 Overview of the ARM applications

Data mining is a process of learning interesting knowledge and generating new information from large datasets. Data mining methods are used for descriptive analysis that describe the relationships among factors as opposed to prescriptive techniques that prescribe the systematic solution for social problems.

The main purpose of this paper is to find interesting combinations of the UQOL attributes that contribute to the high desirability of cities. Therefore, the method’s capability in finding these combinations was considered when selecting an appropriate model. This paper is the first that has applied ARM in UQOL context.

ARM was introduced by Agrawal et al. in 1993 [83] and is a popular and powerful data mining technique that is capable of discovering interesting relationships in a dataset including frequent patterns, associations, and co-occurrences. ARM was first used in marketing science to analyze marketing activities and offer promotions, accordingly. The model was originated from mining a market-basket dataset to find interesting linkages between product items [83]. The goal was to explore the rules that could predict the occurrence of an item (e.g., purchasing soda) in the transactions according to the occurrence of other items (e.g., purchasing pizza). The usefulness of discovering interesting rules between variables has led to applying association rule-based models in many different areas of research. In transportation engineering, it was used to investigate sets of errors associated with the occurrence of defects in traffic sensors [84] as well as in analyzing hazardous materials traffic crashes [85]. In healthcare, it has been utilized in medical diagnosis to investigate the sets of sick and healthy factors that are associated with heart disease [86], and in the banking sector, this method has been leveraged to detect what sets of factors in credit transactions could lead to fraud [87] and characterize the profitable groups of customers [88]. In this paper, association rule mining is utilized in the municipal quality-of-life context to uncover novel, implicit, and previously unknown relationships between the UQOL attributes and desirability of cities.

4.2 The ARM definitions

In this section, we employed the ARM algorithm to identify sets of UQOL attributes that are strongly associated with high levels of desirability in the largest U.S. cities. The ARM model provides a comprehensive evaluation of the quality-of-life indicators, enabling us to explore the complex and often unknown interrelationships between attributes that contribute to high desirability.

To facilitate a clear understanding of the ARM model, we provide definitions and notation commonly used in literature.

4.3 ARM definitions

The following formal notation and definitions are used in the ARM model [89]:

  • Set of items: I = {I1, I2,:::, In}, which represents the set of UQOL variables in this study.

  • Set of transactions: D = {T1, T2,:::, Tm}, which is the task-relevant database, i.e., the set of large cities in this study.

  • Transaction T is a set of items such that T ⊆ I; each transaction is associated with an identifier, TID.

  • An association rule has the form “if X, then likely Y”, denoted as X ⇒ Y, where X and Y are itemsets. X is called the antecedent or left-hand-side (LHS), while Y is called the consequence or right-hand-side (RHS), with X ⊆ I, Y ⊆ I, and X ⋂ Y = ∅.

  • The support of a rule X ⇒ Y in the transaction set D is the number of transactions in D with both itemsets of X and Y to the total number of transactions in the database. Support is calculated as:

    $$Support \, (X \Rightarrow Y) = P \, (X \cap Y)$$
    (2)

  • The confidence of a rule X ⇒ Y is the percentage of transactions in D with both itemsets of X and Y to the total number of transactions that contain itemset X. Confidence is calculated as:

    $$Confidence \, (X \Rightarrow Y)\, = \,P\left( {Y|X} \right) \, = \frac{P(X \cap Y)}{{P(X)}}$$
    (3)

  • The goal of ARM is to find all the rules that satisfy a user-specified minimum support threshold (min_supp) and a minimum confidence threshold (min_conf), which help exclude uninteresting rules [89, 90]. These parameters are crucial in determining the strength and relevance of the rules generated. Appropriate setting of these thresholds is one important factor for the reliability of the ARM technique. Setting them too low may result in many rules, some of which might be trivial or coincidental. Conversely, setting them too high might miss out on potentially interesting associations.

  • If the generated rules using support and confidence measures are still not interesting, lift can be used as another measure for improving association rules. Lift is a measure of correlation and is calculated as:

    $$Lift \, (X \Rightarrow Y) = \frac{{P({\text{Y }}|{\text{ X}}){ }}}{P\left( X \right)} \; = \;\frac{P(X \cap Y)}{{P(X) P(Y)}}$$
    (4)

  • This measures the increase in probability of Y if X occurs. If Lift (X ⇒ Y) > 1, X and Y are correlated, while if Lift (X ⇒ Y) = 1, X and Y are independent.

4.4 Apriori algorithm

The Apriori algorithm is a commonly used technique for association rule mining in databases [91, 92]. It utilizes the Apriori pruning principle, which states that any infrequent set of items must have all of its supersets to be infrequent as well. The algorithm uses a level-wise search, where k-itemsets are used to explore (k + 1)-itemsets and determine the strength of the connection between the itemsets of X and Y. In our study, we identified that {education = above-the-average ⇒ Desirability = High} was infrequent, and found that {education = above-the-average, crime.rate = above-the-average ⇒ Desirability = High} was either equally or less infrequent. For more details on the Apriori principle, refer to Agrawal and Srikant [89].

5 Results and discussions

This section presents the findings of the study on using the ARM model for exploring urban desirability. The study investigates the rules that identify sets of UQOL attributes associated with high levels of desirability in the largest U.S. cities. The results demonstrate that a comprehensive evaluation of city desirability requires considering multiple factors rather than a single-level assessment. Additionally, the study applies the concept of order-qualifiers and order-winners from operations management in the municipal context to gain insights into enhancing the desirability of cities.

5.1 Association rules

The ARM technique is designed to identify factors that collectively predict the occurrence of an event. This study specifically aims to identify antecedent UQOL factors (X) associated with high desirability of cities (Y). The study employs the ARULES package [66] to derive rules with a consequent limited to high desirability. Frequent itemsets were generated using the Apriori algorithm, and only association rules exceeding the minimum confidence and support thresholds were included in the results. The study identified interesting rules using min_supp = 3 and min_conf = 1. The Support, Confidence, and Lift metrics for the top 10 rules are presented in Fig. 1.

Fig. 1
figure 1

Scatter Plot for Association Rules

Full size image

Table 2 provides additional information about the association rule set, which is categorized based on the number of items found in each rule. The results indicate that the rules predicting high desirability of cities are highly correlated and dependent on the attributes being investigated. This is demonstrated by the high Lift (X ⇒ Y) values, which are all 9.9, indicating a strong correlation between X and Y. Additionally, the Confidence (X ⇒ Y) values are all 100%, meaning that whenever X occurs, Y and only Y occur as a consequence.

Table 2 Association Rule Set
Full size table

The Support (X ⇒ Y) values for the 4-, 5-, and 6-item association rules are 4.04 or 5.05, demonstrating a high level of support given that Y was narrowed down to a small number of cities with a desirability score above eight. The 2-item association rule has a Support value of 3.03, which is also considered high given the limited number of cities with top-quality characteristics. Overall, these results provide strong evidence that the occurrence of certain attributes is highly predictive of high desirability in cities.

5.2 Combined UQOL index

The Combined UQOL index reveals the interdependent relationships between UQOL attributes that contribute to highly desirable cities. Our findings, presented in Table 2, demonstrate that no single UQOL attribute alone predicts high desirability. Instead, a combination of attributes is necessary for a city to achieve high desirability. Notably, nine out of ten rules require at least four attributes to occur together to achieve a highly desirable city (Y). This suggests that high desirability is the result of the occurrence of several factors working together. The confidence threshold of 100% used in the model confirms the strength of these interrelationships. This implies that not only does the combination of UQOL attributes found in the rules lead to high desirability, but there is also no combination that results in an alternative outcome. For example, rule (1) shows the following relationship:

Confidence (Median Salary = above-the-average, Wellbeing = above-the-average, Median Household Income = above-the-average, & Cost of Living = below-the-average ⇒ Desirability = High) = 1.

In our dataset, this rule indicates that every large city with values above the average for median salary, community wellbeing, median household income, and below the average for cost of living is perceived as highly desirable (rated above 8) by its citizens. This is demonstrated by the probability of high desirability given these items: P (Desirability = High|Median Salary = above-the-average, Wellbeing = above-the-average, Median Household Income = above-the-average, Cost of Living = below-the-average) = 1. The same interpretation applies to all the rules identified in this study, given that the minimum confidence threshold was set to 1.

5.3 UQOL order-qualifiers and order-winners

Upon analyzing the patterns of items in association rules, it was observed that certain factors are critical for a city to be considered highly desirable. Specifically, in 4-, 5-, and 6-item association rules, cities with above-average median salary and community wellbeing attributes were found to be highly desirable. In fact, no highly desirable city was found without being qualified in providing these two characteristics. Thus, median salary and community wellbeing attributes are known as UQOL order-qualifiers, and they are essential for a city to be competitive in terms of desirability. Controversial findings have emerged regarding the influence of personal income on city desirability and overall happiness. Nevertheless, our study underscores the significance of personal salary as a crucial factor contributing to city desirability across major cities in the United States.

Furthermore, frequent patterns of items in association rules showed that median household income, cost of living, unemployment rate, and commute time are also important factors. To achieve high desirability, at least two of these factors must exist alongside the UQOL order-qualifiers. The results of the 4-itemset association rules indicate that any combination of two factors is satisfactory, except for the combination of unemployment rate and cost of living, which requires the city to be better than average in either median household income or commute time. Similarly, the combination of commute time and median household income requires the addition of either cost of living or unemployment rate. The 6-itemset association rule indicates that the combination of all six attributes leads to high desirability. However, this rule does not provide significant additional value compared to the 4- and 5-itemset rules.

The top 10 rules and the frequency of occurrence for each item can be found in Fig. 2, which illustrates how different combinations of factors can lead to high desirability on the RHS. The thickness of the lines represents the support of the rules.

Fig. 2
figure 2

Visualization of the Top 10 Association Rules using Parallel Coordinates Plot

Full size image

As depicted in Fig. 2, the model has identified a rule (rule #10 in Table 2) that does not include the two UQOL order-qualifiers yet still leads to high desirability:

(Commute Time = above-the-average & Top-Quality Characteristic = yes ⇒ Desirability = High)

This 2-item rule is a strong rule with Lift = 9.9, indicating a very high correlation between X and Y, and Confidence = 1, indicating that every time itemset X has occurred, it has led to Y and only Y. While this rule does not demonstrate that having a top-quality characteristic alone result in high desirability, it highlights the importance of being well-known for a specific characteristic in a city. Our findings suggest that only large cities with above-average commute time and a top-quality characteristic are perceived as highly desirable by citizens (i.e., UQOL order-winners). This suggests that being recognized for a quality could be significant for certain categories of cities, such as those that are more populous and denser.

It is important to note that the concept of UQOL order-winners and qualifiers in the context of municipal quality-of-life is subject to change as nations advance and residents’ preferences evolve. Therefore, further investigation of UQOL order-winner concept by future UQOL scholars is necessary to gain a better understanding of this phenomenon.

In summary, the application of ARM in UQOL research represents a shift towards more data-driven, empirical approaches in urban studies. This model adeptly handles both subjective and objective data inputs. ARM is particularly effective at uncovering hidden patterns and relationships between various UQOL factors. Unlike traditional statistical methods that often focus on linear relationships, ARM can reveal more complex associations that might otherwise go unnoticed. Given that UQOL is a multidimensional concept involving various factors, ARM's ability to analyze multiple dimensions concurrently offers a more holistic understanding of what contributes to the quality of urban life. Furthermore, ARM can be used to predict future trends in UQOL based on current and historical data. This predictive power is crucial for proactive urban planning and management. The rules and patterns discovered through ARM can inform urban planners and policymakers. By understanding the intricate relationships between different UQOL factors, more effective and targeted urban development strategies can be formulated.

However, it's important to recognize that while ARM is excellent for identifying associations between variables, it does not establish causality. This means that while it can highlight relationships, it cannot confirm that one factor causes another. Also, it should be noted that the rules discovered in one urban context may not be applicable to another due to different socio-economic, cultural, and environmental factors.

6 Conclusion and future study

Based on the results of our study, it is evident that the quality of life in urban areas is a complex and multidimensional concept that should be taken into consideration by urban governors. Our study makes a significant contribution to the understanding of UQOL by addressing a notable gap in existing literature. Prior research has often overlooked the intricate interdependencies among UQOL attributes and their collective impact on citizen satisfaction. Our application of the ARM algorithm has allowed us to uncover previously unknown interrelationships between various factors that contribute to a city's perceived desirability. Furthermore, previous research has not adequately explored the factors that constitute minimum requirements for a city to be considered desirable, nor has it addressed whether possessing a nationally recognized top-quality characteristic contributes to a city's attractiveness. Through our analysis, we have identified some key factors that consistently appear in rules associated with high desirability of major cities in the United States, including median salary and community wellbeing. They alone do not lead to the city’s desirability; however, they are indispensable factors for ensuring the satisfaction of city residents. These factors can be considered urban desirability qualifiers, as they determine the minimum requirements a city must meet to be considered highly desirable.

Our research aligns with the findings of Marques et al. discussing the importance of maintaining a relationship with nature for human health and wellbeing, especially in urban areas. Their study shows how architecture and landscape architecture, and overall community wellbeing can transform urban environments into desirable places to live [93]. Khorshidifard also discusses the concept of mutual aid and its role in building wellbeing and equity in communities. It highlights a project in Springfield that addresses food insecurity, indicating the importance of community wellbeing in urban areas [94]. A study by Ray analyzes the distribution of pre-tax wages and salaries in the ten largest metropolitan areas of the USA. It highlights significant differences in average salaries across these areas, pointing to the impact of regional variations on labor market conditions and, by extension, on city desirability [95].

It is imperative to acknowledge that these factors exhibit variations across diverse geographical regions worldwide. Therefore, future studies are suggested to incorporate distinctions in cultural dynamics and other relevant criteria when delineating the qualifiers for urban desirability within varying countries. Another limitation of this study is the lack of spatial configuration factors such as walkability and pedestrian and cyclist-friendly design in our set of variables. While we have utilized the community wellbeing variable to measure accessibility to amenities in various areas, we recognize that the specific physical layout and design of urban spaces play a crucial role in determining a city's livability and functionality. Future research could benefit from incorporating these factors to achieve a more comprehensive understanding of the UQOL index.

Our findings also highlight the importance of taking a comprehensive approach to improving the quality of life in urban areas. By addressing multiple factors simultaneously, urban governors can create a more holistic and sustainable environment that fosters the satisfaction of its citizens. It is important to note that the quality of life in urban areas is not a one-size-fits-all concept, as different cities may have unique characteristics that contribute to their perceived desirability. Accordingly, our study examined whether the presence of a nationwide top-quality factor, where the city's overall quality surpasses that of other cities in the country, confer a competitive advantage in being perceived as highly desirable. Our results demonstrate that it can significantly contribute to a city's perceived desirability in densely populated urban areas. This implies that in more crowded cities, possessing a distinguishing characteristic that elevates the city's reputation in this aspect can attract residents and positively influence their perception of desirability associated with residing in that area. Our study concentrated on large cities. It is recommended that future study explore whether the existence of top-quality characteristics among small and medium-sized cities is associated with high desirability.

In conclusion, our study underscores the need for urban governors to adopt a multidimensional approach to improving citizens’ satisfaction in urban areas. By understanding the complex interplay between various factors, they can create a more desirable and sustainable environment for their citizens.