(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Developing a transit desert interactive dashboard: Supervised modeling for forecasting transit deserts [1]

['Seung Jun Choi', 'Urban Information Lab', 'The School Of Architecture', 'The University Of Texas At Austin', 'Austin', 'Tx', 'United States Of America', 'Junfeng Jiao']

Date: 2024-08

Abstract Transit deserts refer to regions with a gap in transit services, with the demand for transit exceeding the supply. This study goes beyond merely identifying transit deserts to suggest actionable solutions. Using a multi-class supervised machine learning framework, we analyzed factors leading to transit deserts, distinguishing demand by gender. Our focus was on peak-time periods. After assessing the Support Vector Machine, Decision Tree, Random Forest, and K-nearest Neighbor, we settled on the Random Forest method, supported by Diverse Counterfactual Explanation and SHapley Additive Explanation in our analysis. The ranking of feature importance in the trained Random Forest model revealed that factors such as density, design, distance to transit, diversity in the built environment, and sociodemographic characteristics significantly contribute to the classification of transit deserts. Diverse Counterfactual Explanation suggested that a reduction in population density and an increase in the proportion of green open spaces would likely facilitate the transformation of transit deserts into transit oases. SHapley Additive Explanation highlighted the differential impact of various features on each identified transit desert. Our analysis results indicate that identifying transit deserts can vary depending on whether the data is aggregated or separated by demographics. We found areas that have unique transit needs based on gender. The disparity in transit services was particularly pronounced for women. Our model pinpointed the core elements that define a transit desert. Broadly, to address transit deserts, strategies should prioritize the needs of disadvantaged groups and enhance the design and accessibility of transit in the built environment. Our research extends existing analyses of transit deserts by leveraging machine learning to develop a predictive model. We developed a machine learning-powered interactive dashboard. Integrating participatory planning approaches with the development of an interactive interface could enhance ongoing community engagement. Planning practices can evolve with AI in the loop.

Citation: Choi SJ, Jiao J (2024) Developing a transit desert interactive dashboard: Supervised modeling for forecasting transit deserts. PLoS ONE 19(7): e0306782. https://doi.org/10.1371/journal.pone.0306782 Editor: Xiao-Dong Yang, Ningbo University, CHINA Received: November 14, 2023; Accepted: June 20, 2024; Published: July 24, 2024 Copyright: © 2024 Choi, Jiao. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: The data used for the analysis is available from the ICPSR public repository at https://deposit.icpsr.umich.edu/deposit/claimResource?tenant=openicpsr&claimId=128244For. Funding: This research was supported by National Science Foundation (NSF-2125858; 2133302; 1952193), the UT Good System Grand Challenge (Good Systems), and the USDOT Cooperative Mobility for Competitive Megaregions University Transportation Center at The University of Texas at Austin (USDOT CM2). Competing interests: The authors declare that they have no known competing interests or personal relationships that could have appeared to influence the work reported in this paper.

I. Introduction The creation of transit gaps with urban expansion has interested transportation planners and has been a focal point of research. Transit gaps generally occur due to the mismatch between demand and supply in the level of services [1]. Notions of “accessibility” to certain transportation uses and “capability” to use them are widely considered to quantify these issues, but analytical outcomes hugely depend on the employed materials and methods [2]. Oftentimes, the research addresses the opportunities and burdens of the “disadvantaged” communities, which generally falls into the transportation equity literature in the transportation field [3]. Historically segregated individuals and communities, including people of color and low-income households [4–6], are typically found to experience significant gaps [3]. The concept of a transit desert, which addresses the quantification of transit gaps, is similar to the concept of a food desert [1,7]. Food deserts (first termed by Cummins and Macinetyre [8]) measure relative access to nutritious foods. Jiao and Dillivan [1] applied this concept to measure the gap in mass transportation systems and coined the term ’transit desert.’ Here, a transit desert refers to a geographical location where a transit-dependent population experiences a shortage in supply to meet their demand [1]. The orientation of the transit desert is tamed by historical planning policies and practices [9]. Existing studies in transit deserts have well investigated its existence around the globe [10–12]. However, they stay at the stage of identifying transit deserts and fail to examine the underlying cause of the resulting transit deserts. It is already known that transit deserts possess a greater number of disadvantaged populations than regions where transit supply surpasses demand [12]. Still, the studies hardly address the cause of transit deserts or reform an ambiguous dialogue that there’s an issue with the supply side [13]. The results of transit desert analysis can be applied to leverage policy practices, but studies remain of a reformist orientation. A structural cause should be addressed to “transform” the practice rather than staying to “reform” the practice [2]. In the meantime, the capabilities of using transportation are different between men and women [7,14]. Women and gender minorities are prone to experiencing more challenges than men when using the transit system because they are more vulnerable to harassment [15]. The need to consider fundamental differences between the two sexes is clear, but often, transportation analysis is akin to using simplified aggregation [16]. The limitation of previous analyses on transit deserts repeats the tradition of using simplified aggregation. The calculation of transit deserts involves classifying multiple transportation factors into demand and supply factors [1]. Both demand and supply are closely tied to the 5D attributes—Design, Density, Diversity, Destination, and Distance—in the built environment. The built environment has been known to significantly impact travel behavior and mode choices [17–21]. Leveraging the relative magnitude of demand and supply, a transit desert is defined as a location where demand is greater than supply. Conversely, a transit oasis is a term used in cases where supply surpasses demand. Transit deserts primarily occur in downtown areas or central business districts, where major transit stations are located [1]. However, their distribution and occurrence vary across types of land use [22], which may be the result of historical land use regulation or policy practice. The practice of segregation policies, like redlining practices when interstate highways were built, has resulted in the urban disparities observed today [23]. The occurrence of a transit desert in this scenario would be the result of structurally segregating the transit-dependent population. Studies on transit deserts have centered on major cities in the United States and European countries [1,11,22]. Lee et al. [12] extended the work by identifying transit deserts in Seoul, South Korea (S. Korea), while integrating the uniqueness of their city. By utilizing real-time floating population data that moves from place to place, they confirmed that transit deserts occur spatiotemporally in real-time. The disadvantaged population, such as low-income households receiving vouchers, was two times higher in transit deserts than in transit oases; the number of people with disabilities was also greater in transit deserts. Transit oases statistically possess more transit infrastructure than transit deserts. The identification of transit deserts aligns with investigating equity concerns in transportation and addressing the challenges of different population groups [24]. Nevertheless, studies on transit deserts remain in the identification stage [1,11–13]. The underlying causes of transit deserts remain unexplored. They reform the notion of equity in transportation rather than transform it [2]. The notion of equity first appeared during the Civil Rights Act of 1964 [25]. In the planning field, equity touches on the issue of distributive justice, as described by Rawls [26], incorporating concepts of equality and addressing the needs, demands, preferences, and willingness of people [27]. It first influenced the conceptualization of climate/environmental justice [28]. Later, reflecting on works in environmental justice concerning the disproportionate burden, transportation scholars coined the term equity in transportation. Transportation equity addresses the benefits and burdens of using transportation across different socio-demographic groups and discusses their relativeness [3]. Measuring transportation equity exists in various forms, and the results depend on the metric we use [2]. The identification of equity; “of what,” “for whom,” and “how much” is addressed in stages [3]. Generally, historically disadvantaged communities, including low-income households and communities of color, are considered [3,29]. Studies often focus on metropolitan areas, and attempts to address conditions in rural or suburban regions are being conducted [2]. The transportation equity “of whom” is likely to diverge when applying the lens of perspectives from different genders. For feminist scholars, the simple aggregation of travel behavior, without considering the uniqueness between sexes and attempts to disaggregate behavior, is pointed out to be lacking in transportation research [16]. Women’s travel behavior differs from that of men [30]. Gender minorities, including women and LGBTQ+ communities, are more prone to harassment [15]. Their safety and security impact their perceptions and actions in cities [31,32]. A threshold definition is preferably applied to address the issues of equity [2,33]. Other approaches include using a graduated scale, index, and Geographic Information System (GIS) integration [2]. The identification of transit deserts relates to previous approaches, respectively. The transit desert evaluates the accessibility of transit, as the number of certain transit stations is included in the supply factors. Measuring accessibility to specific transportation infrastructure or programs is a key concern in transportation equity [34], which discusses different human capabilities in using transportation [29]. Transit deserts use the aggregation of multiple variables in regions. The setting would be more nuanced if the measurement of capabilities were more oriented toward regional capabilities. Overall, the transportation equity framework is quite solid and effective at addressing issues in transportation. However, some scholars have critiqued that there’s a habit of perpetuating tradition in the form of paying lip service to the notion of equity [3,35]. As an alternative, Karner et al. [3] introduced a "justice" framing, suggesting a disentanglement of planning practices with a society-centric oriented practice. Community leaders and local grassroots organizations are suggested as stakeholders to truly transform transportation planning practices. The manner and extent to which planning practitioners communicate with local stakeholders are critical to actualizing the notion of justice. Our analysis of transit deserts is rooted in transportation equity. Given that transit deserts address the burdens of disadvantaged groups, they consider those groups’ transportation needs by quantifying regional capabilities. We expand the work by disaggregating the demand factors according to different sexes to broaden the equity “of whom” in transit deserts. Moreover, we investigate the underlying causes of transit deserts to suggest actual means to mitigate them. We present a supervised modeling framework for forecasting transit deserts to examine factors associated with the cause of transit deserts with the validation of our presented model. Recently identified characteristics of transit deserts are applied. Our studies propose ways to utilize the investigation of transit deserts to inform communities, ensuring that the analysis of transit deserts and transportation equity is genuinely used for practical purposes in the field, moving from equity to justice in the future. We leverage implications usable so that practitioners get a glimpse of not only identifying transit deserts but actually making use of them. In summary, the limitations of existing studies are as follows: Transit desert analyses have largely stayed at the identification stage, repeating a reformist orientation. The causes of transit deserts remain unexplored. There is uncertainty about how identification from transit desert analyses can be effectively applied in planning practices. Transportation equity literature, including transit desert analysis, often relies on aggregated data, failing to clarify the differences in travel behavior between men and women. The advancement of machine learning techniques, combined with easy access to emerging empirical data sources, can bridge the existing knowledge gap. We synthesize these elements through the following research contributions: We integrated transit desert analysis with multi-class supervised machine learning to demonstrate potential planning interventions through forecasting and the creation of an interactive dashboard.

In the identification of transit deserts, we disaggregated the demand factor by gender, comparing men and women, to illustrate how results differ from conventional aggregation.

We investigated the causes of transit deserts using feature estimation modules, suggesting alternatives for modifying transportation supply factors, including attributes of the built environment. The remainder of this article is organized as follows. First, our study presents employed materials and methods. Then, the analysis results are presented. They are further discussed with a demonstration of the interactive dashboard for planning practice. Lastly, we conclude by leaving out the limitations of our study and directions for future research.

IV. Discussion This study introduced a supervised machine learning approach to predicting transit deserts. While primarily drawing from the research that explored the spatiotemporal identification of transit deserts in Seoul by Lee et al. [12], we examined transit deserts during peak periods. We employed a backpropagation approach for output classification. Both aggregated demand and gender-disaggregated demand were considered during the sample validation process. Our primary objective was to progress beyond just identifying transit deserts. We aimed to bridge the gap from identification to mitigation, linking transit desert analysis with transportation equity and challenging the oversimplifications commonly found in transportation equity literature. The findings of our study are as follows. First, analysis results concerning transportation equity differ when using aggregated versus disaggregated data. When disentangling transit demand factors according to different sexes, cases of transit deserts and transit oases only appear in specific scenarios. It’s important to note that both transit deserts and transit oases signify imbalances between transit demand and supply. Women experience a greater mismatch, evident from a higher number of transit deserts and transit oases compared to the results from aggregated data and men’s demand. The differences highlight that transportation analysis should account for the fundamental differences in travel behavior among individuals [16,30], in addition to the variations observed when different methodologies are employed [2]. Through the lens of equity, it’s imperative to understand that the capability and availability of transportation vary among individuals. Commonly, women and gender minorities face harassment while using public transit [15,51]. Planning practices aimed at reducing disparities among specific community groups should steer clear of oversimplifying scenarios through aggregation. Simple normative assumptions about race, gender, and sexuality should not shape urban planning [52]. As evidenced in the characteristics of unique transit deserts compared to aggregated cases (refer to Fig 8), planning policies should aim to address distinct existing needs. While it is crucial to prioritize the requirements of disadvantaged groups, efforts to mitigate transit deserts should be tailored to contemporary conditions. Such measures would inherently differ when enhancing the factors of transit supply. Second, we successfully modeled the classification of transit deserts and transit oases using a supervised framework. The emergence of AI-affiliated toolkits, including machine learning and deep learning, has garnered significant interest among researchers. The challenge now is determining the best ways to utilize new methods. Our research sought to merge a branch of transportation theory with new methodologies. In particular, we integrated transportation equity literature with the concept of transit deserts. We compared multiple multi-class classification machine learning models and found the RF model outperforming the rest. The selected RF was notably effective in identifying regions as transit deserts, N/A, or transit oases (refer to Figs 2 and 6). Furthermore, the model captured the characteristics of transit deserts and transit oases in Seoul. As depicted in Fig 7, supply dimensions were closely tied to transit oases, whereas the presence of disadvantaged groups correlated strongly with the likelihood of regions being classified as transit deserts [12]. A potential application of our model is to integrate it into a dashboard or interactive toolkit, with the model functioning in the background. For instance, pre-trained models can be saved in a.sav format and later imported using the pickle module in a Python environment. Such an approach can facilitate the development of an interactive dashboard, as illustrated in Fig 9. Our dashboard demonstration offers a simplified method for altering the numeric values of significant variables. As users modify these inputs, the classification result updates in real-time. To illustrate, comparing the classification result labeled as N/A in Fig 9A. with Fig 9B. shows that increasing the number of disadvantaged individuals prompts the model to reclassify the region as a transit desert. Drawing parallels with the crowdsourcing toolkit [53], using this model in conjunction with community outreach, engagement, and participatory planning practices could allow residents to better understand and influence the characteristics of their communities. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 9. (a) Demonstration of the transit desert dashboard displaying an N/A case; (b) Demonstration of the transit desert dashboard displaying a transit desert case. https://doi.org/10.1371/journal.pone.0306782.g009 Tracing back through planning history, the evolution of equity planning has been a gradual response to societal and economic factors [54]. Its origins can be traced to advocacy planning in the 1960s and 1970s, which marked a pivotal shift from the physical realm of planning towards addressing pressing social and economic issues, such as urban poverty and unemployment [54]. The history of segregation, characterized by redlining and exclusionary zoning, marginalized low-income working-class individuals and people of color. This necessitated a shift where planners should advocate for marginalized communities in pursuit of public goods. Equity planning has evolved gradually, influenced by historical events and broader societal and economic factors [54]. Reece [54] well documented the evolution of equity planning. During the Progressive Era, equity planning responded to issues such as immigration, industrialization, the emergence of tenements, and increasing inequality. The New Deal era prompted responses to address the Great Depression. Advocacy planning emerged in response to the Civil Rights Movement, the legacy of urban renewal, and social conflict. The Just City era addresses challenges related to globalization, immigration, rising inequality, and gentrification. Given that equity planning aims to pursue social justice and advocate for the needs of the marginalized, its core essence remains consistent. Transportation equity aligns with traditional equity planning paradigms, where the goal is an equitable redistribution of transportation resources. However, a critical perspective suggests that advocacy planners, despite their intentions, may exhibit political naivety and engage in tokenistic practices without yielding actual outcomes. This has led some scholars to critique the employment of ’equity’ as mere lip service [3,35], without effecting meaningful changes in the decision-making process and its deliberations. A foundational case in equity planning from Cleveland by Krumholz [55] suggested that planners should leverage the power of information, analysis, and insight with an equity-focused lens in the decision-making process. However, the challenge lies in overcoming the status quo and ensuring the planning process remains truly informative. Emphasizing community participation and striving to be genuinely ’informative’ are identified as critical to the success of equity planning. With the advent of AI and related technologies, equity planning is poised for the next evolutionary leap. Our model introduces AI-augmented planning with a ’planner-on-the-loop’ approach [56]. Here, the focus remains on planners maintaining a strong lens of equity in planning. The integration of AI, transitioning from an ’AI out of the loop’ to an ’AI in the loop’ approach, refines planning practices and policies. Our empirical approach exemplifies how, in addressing issues such as transit deserts, transportation equity planning can progress to a more advanced stage while incorporating classic planning principles. John Forester [57], who has significantly influenced contemporary planning practices with his communication action theory, highlights information as a source of power. He asserts the planner’s role in addressing information and anticipating misinformation while communicating with various stakeholders in the decision-making process. Communicative planners emphasize about crucial role in facilitating collaboration among stakeholders and affected groups through a creative process [58]. Their tools include various communication practices, such as listening, storytelling, rhetoric, mediation, and the use of metaphors. Now, with AI in the loop, contemporary communication practices are at a turning point. An interactive dashboard fundamentally requires human auditing and management. The essence of a planner’s role while utilizing AI depends on how effectively planners communicate using AI or systems incorporating AI. This involves issues such as how we use AI, how we present it, and how it is employed in decision-making processes and outcomes. Historical epistemological debates in planning still resonate today. Based on the theoretical framework of communicative action theorists, perhaps we should explicitly share the limitations of using the interactive dashboard. The fundamental dilemma of using our dashboard likely sustains criticism about biases in the dataset and errors in machine learning models [59]. How people react to decisions made by computational algorithms also varies depending on how they perceive the decision to be, whether less fair or trustworthy [60]. Planners, acting as facilitators or negotiators in the communication process, should strive to give equal opportunities to both speakers and listeners. If planners or developers are perceived as having more power or authority because they oversee the algorithmic decision-making system, we should then attempt to mitigate our inherent authority. Fjeld et al. [61] described accountability in AI as the anticipation that those who design, develop, and deploy AI systems will adhere to established standards and laws, ensuring AI operates correctly throughout its lifespan. Thus, the issue of AI systems compromising accountability lies with those who designed them. Situating value in the AI system needs further study. Lastly, transit deserts continue to persist in Seoul. Due to changing travel behaviors, population movements, and the implementation of new planning policies, mismatches between transit demand and supply inevitably occur. If we analyze the identified transit deserts based on data primarily obtained from the Seoul Metropolitan Government database, transit deserts are areas where disadvantaged groups are more likely to reside. The result is similar to the transit desert analysis in Seoul conducted by Lee et al. [12]. The average number of elderly living alone, low-income households, and residents with disabilities in the transit deserts was 1,498, 1,172, and 1,191, respectively. These figures are higher than those in transit oases and regions that are neither. In transit oases, the averages for elderly living alone, low-income households, and residents with disabilities were 978, 659, and 896, respectively. For regions neither classified as transit deserts nor oases, these numbers were 832, 580, and 798, respectively. In the meantime, the average population density in transit deserts was the highest at 33,237 per km2, compared to 9,074 per km2 in transit oases and 6,445 per km2 in regions neither classified as transit deserts nor oases. Throughout the study period, the average hourly numbers of metro, bus, and bike users in transit deserts were 48,241, 8,735, and 368, respectively. These figures are lower than those in transit oases, where the average hourly numbers for metro, bus, and bike users were 48,999, 9,074, and 713, respectively. Regions classified as neither transit deserts nor oases reported even lower mean values: 21,137 for metro, 6,445 for bus, and 309 for bike users. It was noted that the destination and distance to transit attributes of the built environment in transit deserts were insufficient compared to those in transit oases. On average, transit deserts had 44 total transit stops, 12 city buses, 145 taxis, 7 bike stations, 36 bus stations, and one metro station. In contrast, transit oases comprised 67 transit options: 62 buses, 570 taxis, 15 bike stations, 50 bus stations, and 2 metro stations. These findings suggest that public transit use in transit deserts is significant and is relatively comparable to use in transit oases. However, the destination and distance attributes of the built environment in transit deserts are less sufficient than those of transit oases. Moreover, transit deserts had the lowest average green open space at 22%, compared to 56% in transit oases and 32% in areas classified as neither. Using disaggregated demand cases, two distinct transit deserts were identified for both men and women. In the former case, the Amsa 1 and Daechi 2 administrative boundaries (locally termed "Dong") were noted. In the latter case, the Seonghyeon and Jamsil 3 administrative boundaries fall into this category. The Amsa 1 administrative boundary reported a relatively high residential ratio of 42%. This region has only 4 bike stations and 13 bus stations and lacks a metro station, which is below average for typical transit desert cases. However, its population density is notably high at 46,137 people per km2. It also has a significant number of disadvantaged groups, with 1,714 elderly living alone, 1,351 low-income households, and 1,692 residents with disabilities. These figures exceed the average characteristics found in transit deserts. On the other hand, the Daechi 2 administrative boundary appears to have more family households with children, as the population aged 10 to 19 makes up 23%, and those aged 40 to 59 constitute 49% of its residents. This area has fewer disadvantaged community members, with relatively low numbers of elderly living alone (594), low-income households (118), and residents with disabilities (604). The Seonhyeon administrative boundary exhibited a number of disadvantaged groups comparable to those in the Amsa administrative boundary. It reported 1,403 elderly living alone, 1,204 low-income households, and 1,436 residents with disabilities. The Seonhyeon administrative boundary has limited public transportation options, with only 4 city buses, 1 bike station, and no metro station. On the other hand, the Jamsil 3 administrative boundary is known for its high population density, at 55,929 per km2. It’s notable for recreational activities, with one of Seoul’s tallest buildings serving as a landmark. This area has 48% green open space and predominantly houses an affluent community. Consequently, there are only 44 low-income households, alongside 728 elderly living alone and 681 residents with disabilities. For general improvements to mitigate the occurrence of transit deserts, enhancing the design and proximity of transit attributes within the built environment can be broadly applied. Specifically, DICE identified that green open spaces, tied to travel satisfaction [62], significantly increase the likelihood of a transit desert transforming into a transit oasis. Transit deserts were also confirmed to have less green open space compared to transit oases, or cases neither classified as deserts nor oases. Improvements to bike route distances, the number of parking lots, bike routes, transit stops, and the installation of additional bus lanes and stops could also play pivotal roles. They have been identified as mitigating factors that reduce the likelihood of regions being classified as transit deserts in SHAP analysis. Additionally, it is crucial to note that communities’ individual characteristics and current statuses should be addressed. For instance, to address the transit needs within the Amsa 1 and Seonhyeon administrative boundaries, it is critical to consider the needs of disadvantaged groups. Public intervention is necessary to enhance the overall public transit services in Amsa 1 and Seonhyeon administrative boundaries. In the Daechi 2 administrative boundary, a transportation policy that caters specifically to the needs of family households appears to be more suitable. The approach for the Jamsil 3 administrative boundary should aim to support recreational trips and effectively manage the area’s comparatively high population density. We are not advocating for a one-size-fits-all planning intervention. Rather, during the decision-making process, when evaluating alternatives with incremental differences, considering the unique characteristics of each region can help planners and policymakers make better decisions. Lastly, transit deserts exist around the globe [1,11,12,22]. The mismatch between transit demand and supply is inevitable. The focus should be not on solving the problem entirely but rather on mitigating the occurrence of transit deserts. The case study of Seoul in S. Korea can be applied to other cities to develop a forecast model to address the causes of transit deserts and advocate for the needs of disadvantaged groups. However, similar to our efforts in addressing localized scenarios, what constitutes transit supply and demand factors should be contextualized. The level of service in transit differs depending on the size of the city. Metropolitan regions, megaregions, urban, suburban, and rural areas have different conditions. The case study of Seoul mainly falls into the category of studying a metropolitan region with a high population density. We do not suggest directly applying the analysis of transit deserts without modifications.

V. Conclusions The present study outlines a supervised research framework for modeling transit desert classification using machine learning. We utilized three class labels: Transit Desert, N/A, Transit Oasis by comparing transportation demand with supply factors. We evaluated several algorithms: SVM, DT, RF, and KNN, and performed hyperparameter optimization. Based on the evaluation metrics, we chose RF. We extracted feature importance from the trained RF model. Both global and local validations of identified transit deserts and the impact of collected features on classification probabilities were analyzed using DICE and SHAP values. Later, we integrated the trained RF model to demonstrate its practical application in developing interactive interface. Our research offers several contributions to the field. Primarily, we expand on existing studies about transit deserts, moving beyond mere investigation to develop a modeling framework that addresses the multifaceted factors associated with transit deserts. Our identification of transit deserts aligns with existing transit desert analyses [12]. Transit deserts are regions where a larger number of disadvantaged groups are located and where the supply of transit is insufficient. Transit desert analysis is a needs gap analysis that quantifies transportation poverty within the scope of transportation equity analysis [63]. However, the challenge lies in the criteria selected for this analysis [63]. It remained uncertain how the results of transit desert analysis might vary should we disaggregate the aggregated transit demand by gender. In our approach to identifying transit deserts, we differentiated the outcomes of transportation equity analyses by comparing aggregated data with disaggregated data (separating demands between men and women). Our study emphasized the importance of contextualizing planning efforts. We found unique transit deserts for both men and women, showing that results change when we look at aggregated data versus disaggregated data. The transit deserts we identified each have their own set of needs. Some require more focus on helping disadvantaged groups, another needs more support for families with children, and another should concentrate on improving options for recreational trips. Moreover, we evaluated multiple machine learning models, culminating in the demonstration of a transit desert dashboard. The interactive dashboard incorporates human intervention into transportation equity analysis. By adjusting the threshold values, users can determine whether their region is identified as a transit desert. Algorithms and AI systems with machine learning become more effective when centered around human needs. Human intervention in the decision-making process helps users trust the outcomes generated by machines [64]. Creating an interactive dashboard differs from a static information viewer dashboard. It opens up new avenues for innovative participatory planning practices. It adds a new tool for communicative action planners, especially. We showcased the case of transportation poverty using a needs-gap analysis approach, but other tools in the planning equity analysis toolkit can also be designed with AI systems. Ultimately, no models are perfect, and each method has its limitations. Holistically combining multiple sets of tools and using them as both means and for outcomes might be beneficial. However, sustained criticism exists about biases and errors in computational algorithms [59]. Future studies should consider exploring the discrepancies between identification by AI and residents’ perceptions in defining their communities, which would validate different use cases. Therefore, qualitative research involving user surveys, interviews, and data auditing is necessary. Future research might also benefit from focusing on specific analyses of green transportation needs. This focus is critical because transportation significantly contributes to greenhouse gas emissions, a major factor in climate change that we need to mitigate. We are increasingly experiencing the impact of climate hazards in our daily lives, leading to the use of the term ’climate extremes’ to describe the severe consequences of climate change. The role of green transportation, such as electrifying transport systems and integrating them with renewable energy, is gaining recognition for its importance. The analysis of transportation equity is linked to environmental justice [3], seeking to fairly address the unequal/disproportionate impacts of emerging issues across different communities. There could be disparities in how green transportation is adopted and accessed. Identifying areas lacking in green transportation, or ’green transit deserts,’ can aid in establishing a baseline that tackles the disparities in the adoption of green transportation. However, it is essential to acknowledge several limitations. Our study primarily focuses on transit deserts during peak-time periods. Supervised machine learning, on its own, can only be optimized by being fine-tuned based on future data insights. While our suggestions for mitigating transit deserts primarily focus on improvements on the supply side of transit, the influence of individual attributes should not be overlooked. Involving machines, or AI, in decision-making requires building trust and situating value in the AI system. This involves enhancing the system’s accountability and mitigating biases in the datasets used for training the machines [59]. Communicative planners hold public meetings, town hall meetings, focus group meetings, charrettes, Delphi methods, and other forums to communicate and build trust. How interactive dashboards can be applied to different types of communicative planning tools remains a critical implementation challenge. We earmark these areas for future research.

[END]
---
[1] Url: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0306782

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.