1. Introduction

2. Literature Review

2.4. Demand Prediction Studies of E-Scooters

Better living conditions, health, education, and employment opportunities are the key contributors to urbanization. Academics and those involved in future design have introduced the smart city idea in relation to all of this [72]. Infrastructure efficiency, operations and services, utilities, energy, governance, lifestyle, business, and mobility in cities can all be improved with the idea of smart cities. Micromobility could enhance the mobility of cities by providing convenient first and last-mile connectivity to public transit [68].

Previous research on shared e-scooter use identified several factors that increase e-scooter demand, including the presence of businesses and industries, transit accessibility, mix of land uses, locations in central business districts, population density, bike scores, student-populated areas, and weather conditions—increased e-scooter demand [3,43,69,70,71].

The use of shared e-scooters [70], biking [72], and other modes of transportation is influenced by land-use and mixed land-use indices [67]. In urban contexts, e-scooter use can strengthen the role of micro-mobility services. For example, the availability of e-scooters in densely populated, walkable regions and their proximity to public transportation hubs may positively impact the increase in public transport ridership.

Tremendous research, including studies [73,74,75] has focused on using machine learning to forecast demand in bike-sharing programs. Specifically, Zhou et al. (6) employed a Markov chain model to predict daily demand for bike-sharing programs [75]. Their findings suggested that the Markov chain model outperformed other models, such as regression, neural network, and support vector machine models.

A graph convolutional network (GCN) model was developed to estimate hourly demand at specific stations within the New York City bike-sharing program [74]. Additionally, multilayer GCN, introduced by [76], was used to capture the spatiotemporal characteristics of bike demand patterns. While the GCN framework shows high predictive performance, it faces limitations due to high processing costs and the inability to incorporate data from newly introduced stations into its predefined graph structure.

This limitation impedes the GCN’s capacity to adapt to real-time changes, especially when new stations are added to the network.

[73] developed an efficient approach to estimate high-resolution e-scooter demand. The newly introduced neural recurrent neural network–decoder ERD framework significantly enhances prediction performance by reducing the mean squared error loss. This model forecasts both the identified and unfulfilled demand effectively.

[7] proposed a fully convolutional network (FCN) model, combined with a masking method and a weighted loss function, termed masked FCN (or MFCN), for predicting dockless e-scooter demand. The model effectively predicted peak time demands and off-peak needs by employing the exponential linear unit (ELU) and the hyperbolic tangent (tanh) activation functions, respectively [77]. This approach was advantageous for examining the larger spatial region of e-scooters. Table 2 provides a summary of demand prediction studies.

[78] utilized the Random Forest method to predict the usage effectiveness of shared electric scooters, employing GPS-based vehicle availability data. The study’s findings highlighted the model’s superior predictive ability, with factors like time of day and the density of Points of Interest (POIs) being highly significant. [79] conducted a study focusing on the use of e-scooters among students and staff on campus. They also employed the Random Forest model, achieving high accuracy in predicting e-scooter usage frequency. These studies, employing various machine learning and deep learning algorithms, effectively incorporate factors such as weather, time, location, and day of the week to accurately predict e-scooter demand in specific areas.

Demand prediction studies for e-scooters significantly enhance service quality and operational planning. These studies primarily focus on predicting spatiotemporal e-scooter demand, facilitating efficient relocation strategies, and informed operational decisions. Various proposed models and techniques, as discussed above, adeptly handle the sparse and fluctuating nature of e-scooter demand data. They enable precise predictions of pick-up and drop-off demands, either hourly or on a daily basis. Such accurate forecasts are crucial for optimizing e-scooter availability and distribution, thus improving user experience and solving first- and last-mile transportation issues. Moreover, identified research gaps in previous studies underscore the importance of integrating regional spatial characteristics and additional variables like weather and social events to further refine prediction accuracy [73,74]. Advanced techniques, including partial dependence plots and feature interactions, are recommended for better interpretability of machine learning models [78]. This paper contributes to the field by addressing these previously noted limitations in e-scooter demand prediction.

Micro-mobility, particularly beneficial for quick urban travel, offers a more environmentally friendly transportation alternative. However, its effectiveness is weather-dependent, suggesting a need for evolution in these modes to provide weather protection. Balancing the need for lightweight, energy-efficient mobility with the challenges of weather adaptability remains a key consideration in the ongoing development of micro-mobility solutions.

Table 2. Overview of Studies on Demand Prediction.

Table 2. Overview of Studies on Demand Prediction.

Reference	Study location	micromobility type	Demand Modelling Methods	Land Use Characteristics Considered (Yes/No)	Weather Conditions Impact Considered (Yes/No)	Days/times of the week/day
[80]	Netherlands	E-bike sharing	Activity Based Travel Demand Model (ABM)	No	No	Yes
[81]	New York City	Bike sharing	Gradient Boosting Models (GBM)	No	Yes	No
[69]	Austin	Shared E-Scooter	Joint Panel Linear Regression (JPLR) Model	Yes	Yes	Yes
[82]	Austin	E-scooter	Gradient Boosting Regression (GBM), Negative Binomial (N.B.) and A Zero-Inflated Negative Binomial Count Model (ZINB)	No	Yes	Yes
[50]	Kelowna, Canada	Shared E-Scooter	Zero-Inflated Negative Binomial (ZINB)	Yes	Yes	Yes
[83]	Istanbul, Turkey	E-Scooters	Artificial Neural Network (ANN)	No	No	No
[51]	Munich, Germany	Shared E-Scooter	Negative Binomial (N.B.) and Consul’sGeneralized Poisson (GP-1) RegressionRandom Forest model	No	Yes	Yes
[52]	Chicago	Shared E-Scooter	Random-Effects Negative Binomial (RENB) Model	Yes	Yes	Yes
[84]	Austin, Texas,	Shared E-Scooters/E-Bikes	Agent-Based Model (ABM)	No	No	No
[85]	China	Bike-Sharing	Generalized Structural Equation Model (GSEM)	Yes	No	Yes
[73]	South Korea	Shared E-Scooter	Encoder–Recurrent Neural Network–Decoder (ERD)	Yes	No	Yes
[86]	New York City	Shared-Bike	Co-Evolving Spatio-TemporalNeural Network (CEST)	Yes	No	Yes
[87]	Shanghai	Shared Bicycles	Integrated Nested Laplace Approximation (INLA)	Yes	No	Yes
[88]	Minneapolis &Louisville, U.S.	Shared E-Scooter	Kernel Density Estimation (KDE)	No	No	Yes
[89]	Taipei	Shared Bike	Spatial Regression Model	Yes	No	Yes
[90]	New York City	Shared Bikes	Zero-Inflated Negative Binomial (ZINB) Model	Yes	Yes	Yes
[91]	Chengdu, China	Shared Bikes	Two-StageRebalancing Model	No	Yes	Yes
[92]	Zu ̈rich, Switzerland	E-bike Sharing	Spatial Regression Model	No	Yes	Yes

Table 2 reveals a significant research gap in e-scooter demand prediction within the Gulf Cooperation Council (GCC) region. Despite the introduction of e-scooters in countries like Qatar, the United Arab Emirates, Saudi Arabia, Kuwait, and Oman, there’s a noticeable lack of focused studies on their usage in these regions.

The challenging hot desert climate of the GCC presents unique obstacles to the use of electric vehicles (EVs), including e-scooters, especially during the hotter months, impacting both usage and performance. This issue is compounded by the limited charging infrastructure [93]. Understanding how e-scooter demand fluctuates during extreme weather and high humidity periods is crucial. Additionally, it’s essential to explore the influence of cultural factors in Arab countries such as Qatar on e-scooter demand and usage. This exploration should encompass societal norms, perceptions of micro-mobility, and potential barriers to adoption. Challenges also arise from the need for dedicated lanes, parking spaces, and city design conducive to e-scooter adoption.

While there is existing research on e-scooter usage globally, the unique environmental and cultural landscape of the GCC necessitates a region-specific analysis. Technological features of e-scooters and user behaviour patterns, such as trip frequency and usage regularity, are likely to exhibit significant differences in the GCC compared to other regions Thus, addressing these underexplored aspects is crucial for a comprehensive understanding of the e-scooter market in the GCC. It also has broader implications for shaping urban mobility policies, informing environmental strategies, and guiding the development of micro-mobility solutions tailored to the specific needs of the GCC region.

3. Methods

3.4. Density-Based Spatial Clustering of Applications with Noise (DBSCAN)

The DBSCAN algorithm operates on the principle that a point $x$ belongs to cluster $c$ if $x$ is close to many points from the cluster $c$. The DBSCAN algorithm requires the setting of two hyperparameters, namely the neighbourhood radius ($\mathit{e}\mathit{p}\mathit{s}$) and the density threshold ($\mathit{m}\mathit{i}\mathit{n}\mathit{P}\mathit{t}\mathit{s}$).

$\mathit{e}\mathit{p}\mathit{s}$: This parameter defines the neighbourhood relationship between any two points $x_1$ and $x_2.$ W $x_1$ and $x_2$ are neighbours if $d(x_1,x_2)\le eps$, where d is the distance between the two points.

$\mathit{m}\mathit{i}\mathit{n}\mathit{P}\mathit{t}\mathit{s}$: This parameter defines a cluster, where a group of points forms a cluster if their count is equal to or greater than $minPts$.

$MinPts$ should be at least equal to the number of dimensions in the dataset, plus one. Thus, when working with a 2-D dataset, $minPts$ is set to 3. Furthermore, $eps$ can be estimated by plotting the kNN distances of the points (i.e., the distance of each point to its k-th nearest neighbour) in decreasing order. A ‘knee’ in the plot indicates an appropriate value for $eps$.

Consequently, using these two parameters DBSCAN classifies data points into one of the following three types as shown in Figure 2:

• Core point: A point is a core point if it has at least $minPts-1$ neighbour.
• Border point: A point is a border point if it has fewer neighbours than minPts, and at least one of these neighbours is a core point.
• Outlier: A point is an outlier if it has fewer neighbours than minPts and none of these neighbours is a core point.

In Figure 2, the parameter minPts is set to 4. Core points, border points, and outlier points are depicted in red, yellow, and blue, respectively

Figure 2. The DBSCAN Classifier. (Source: https://en.wikipedia.org/wiki/DBSCAN).

Figure 2. The DBSCAN Classifier. (Source: https://en.wikipedia.org/wiki/DBSCAN).

DBSCAN steps of solution

1

Set the values of minPts and eps.

2

Randomly select a point $x$ and classify it.

• If $x$ is classified as a core, then its neighbours become members of the new cluster. If $x’s$ neighbours are labelled as core points, then their neighbours become members as well.
• If $x$ is not a core point, label it as a noise and select another random point.

3

Once the cluster cannot be grown further, randomly choose another point that does not belong to any cluster so far and repeat the process.

4

Stop if all points are labelled.

4. Experimental Work

4.8. Daily Aggregated Trip Data Analysis

Due to the fact that there is one data point per day in the daily aggregated dataset, we used the average temperature and humidity of the day as two weather predictors. Other weather predictors include the minimum and maximum temperatures and humidity observed on that day. Moreover, the day and month were included as predictors to account for possible seasonality. Finally, because this daily aggregated data represents a time series of the total number of trips per day, we used the observed trip counts per day $t-k,where1\le k\le 7$, to explain the variability in the observed trips per day $t$.

The daily aggregated data was used to train a Random Forest model. Consequently, the trained model was applied to sort the independent variables/predictors based on their importance. Moreover, we utilized the partial dependence plot to better understand the relationship between the weather variables and the trip counts. To visually inspect the trained Random Forest model’s performance, we plotted the ground truth trip counts versus the predicted trip counts, as illustrated in Figure 14. A linear relationship between the prediction and the ground truth confirmed that the model learned from the data and was valid.

As shown in Figure 15, the previously observed trip counts are ranked at the top among the predictors, which is expectable due to the time dependency between successive trip counts. Additionally, it is observed that the average temperature is the most influential weather predictor, while in general, temperature-related predictors are deemed more important than those related to humidity.

Figure 16 shows the 2-D partial dependence of trip counts on maximum temperature and humidity as well as average temperature and humidity. From the figure, the following findings can be observed:

• The variation in the daily trip count is slight with respect to maximum temperature and humidity.
• The variation in the daily trip count is relatively considerable with respect to average temperature and humidity.
• The topology of the surface of the PDP is almost identical in panels (a) and (b) of Figure 16.
• At a given average temperature, the change in trip counts is small; for example, at high humidity, there is a slight drop in trip counts.
• The trip count increases as the average temperature rises, reaching its peak at around 80 F, after which it decreases with further temperature increases.

Figure 16. The 2-D Partial Dependence Plot of (a) Max Temperature and Humidity and (b) Average Temperature and Humidity.

Figure 16. The 2-D Partial Dependence Plot of (a) Max Temperature and Humidity and (b) Average Temperature and Humidity.

4.9. Hourly Aggregated Trip Data Analysis

In this variation of the trips’ dataset, we have one data point per hour for each of the five clusters. Hence, we utilized detailed weather datasets (i.e., the daily weather dataset) to obtain the weather information every hour. We employed three weather predictors; temperature, humidity and wind speed. To account for any possible seasonality, we included the time of the day, the day of the week and the month as predictors. Moreover, to incorporate time dependency into our model, we used the observed trip counts at hour $t-k,where1\le k\le 7$, to explain the variability in the observed trips at that hour $t$.

The hourly aggregated data for each cluster was used to train a separate Random Forest model. Consequently, we used the trained model to rank the independent variables/predictors based on their importance. Moreover, we employed the partial dependence plot to gain insight into the relationship between the temperature, humidity, wind speed and trip counts

Figure 17. Visual Inspection of Model Quality for Trained Hourly Aggregated Data.

Figure 17. Visual Inspection of Model Quality for Trained Hourly Aggregated Data.

As shown in Figure 18, the previously observed trip counts are ranked as top predictors, a defendable finding due to the time dependency between the successive hourly trip counts. Moreover, temperature is observed to be the most influential weather predictor, followed by humidity as the second most influential, and wind speed as the least influential predictor.

The next step in the analysis is to understand the dependency of the trip counts on the previously observed trip counts. Consequently, the 1-D PDP shown in Figure 19. reveals the following findings,

• There is a linear relationship between the trip count and the observed trip counts of the previous hours.
• The linear relationship reaches saturation (i.e., flattens out) early when the PDP-dependent variable belongs to $\in \{t_3,t_4,t_5.t_6.t_7\}$.
• The variation of the daily trip count is the largest with respect to the observed trip count at $t_1$.
• The variation of the daily trip count gets smaller as the lag increases.

Figure 19. Partial Dependence Plot of the Six Statistically Significant Predictors.

Figure 19. Partial Dependence Plot of the Six Statistically Significant Predictors.

Finally, we estimated the 2-D PDP of the temperature and humidity. The visual inspection of Figure 20. shows the following,

• The hourly trip count increases as the temperature rises.
• The hourly trip count decreases as the humidity increases.

Figure 20. The 2-D Partial Dependence Plot of Temperature and Humidity.

Figure 20. The 2-D Partial Dependence Plot of Temperature and Humidity.

5. Conclusions

• Weather Variations’ Impact on E-scooter Usage (H1): The hypothesis that weather variations, including temperature and humidity, would significantly influence e-scooter trip frequency was strongly supported. The data revealed a clear preference for e-scooter usage during warmer temperatures and lower humidity levels, aligning with the theoretical framework that posits environmental conditions as critical determinants of transportation mode choice. This finding suggests that e-scooter services need to adapt their operational strategies to weather patterns, potentially adjusting availability and marketing strategies according to seasonal variations.
• Influence of Land Use Characteristics on E-scooter Trips (H2): Our analysis confirmed the hypothesis that land use characteristics significantly affect e-scooter trip distributions. High-density areas with abundant employment, educational opportunities, and mixed-use development were associated with increased e-scooter activity. Conversely, the presence of Qatari female students as a demographic factor led to reduced e-scooter usage, indicating cultural and societal norms as critical factors in mobility choices. These insights highlight the importance of considering local demographic and cultural nuances in the planning and deployment of shared mobility solutions.
• Interaction Between Weather Conditions and Land Use on E-scooter Usage (H3): The study explored the moderating effect of land use characteristics on the relationship between weather conditions and e-scooter usage. While the data provided some evidence for this hypothesis, the findings underscore the complexity of interactions between these variables. It suggests that future research should further investigate how specific land use types can mitigate or amplify the effects of adverse weather conditions on shared mobility usage.

6. Broader Implications and Future Directions:

References

1. Smith, C.S. and J.P. Schwieterman, E-scooter scenarios: evaluating the potential mobility benefits of shared dockless scooters in Chicago. 2018.
2. Brown, A., et al., Impeding access: The frequency and characteristics of improper scooter, bike, and car parking. Transportation research interdisciplinary perspectives, 2020. 4: p. 100099.
3. Hosseinzadeh, A., et al., E-scooters and sustainability: Investigating the relationship between the density of E-scooter trips and characteristics of sustainable urban development. Sustainable cities and society, 2021. 66: p. 102624.
4. Hosseinzadeh, A., A. Karimpour, and R. Kluger, Factors influencing shared micromobility services: An analysis of e-scooters and bikeshare. Transportation Research Part D: Transport and Environment, 2021. 100: p. 103047.
5. Mathew, J.K., M. Liu, and D.M. Bullock. Impact of weather on shared electric scooter utilization. in 2019 IEEE Intelligent Transportation Systems Conference (ITSC). 2019. IEEE.
6. Nikiforiadis, A., et al., Analysis of attitudes and engagement of shared e-scooter users. Transportation research part D: transport and environment, 2021. 94: p. 102790.
7. Phithakkitnukooon, S., K. Patanukhom, and M.G. Demissie, Predicting spatiotemporal demand of dockless E-scooter sharing services with a masked fully convolutional network. ISPRS International Journal of Geo-Information, 2021. 10(11): p. 773.
8. James, O., et al., Pedestrians and e-scooters: An initial look at e-scooter parking and perceptions by riders and non-riders. Sustainability, 2019. 11(20): p. 5591.
9. Sanders, R.L., M. Branion-Calles, and T.A. Nelson, To scoot or not to scoot: Findings from a recent survey about the benefits and barriers of using E-scooters for riders and non-riders. Transportation Research Part A: Policy and Practice, 2020. 139: p. 217-227. [CrossRef]
10. Younes, H., et al., Comparing the temporal determinants of dockless scooter-share and station-based bike-share in Washington, DC. Transportation Research Part A: Policy and Practice, 2020. 134: p. 308-320. [CrossRef]
11. NACTO, Shared Micromobility in the U.S.: 2018. 2018, National Association of City Transportation Officials: United States.
12. Chowdhury, A., et al., Shared mobility devices in Arlington. Retrieved from Arlington: https://ralphbu. files. wordpress. com/2019/05/working-draft-report_051419-1. pdf, 2019.
13. Laa, B. and U. Leth, Survey of E-scooter users in Vienna: Who they are and how they ride. Journal of transport geography, 2020. 89: p. 102874.
14. Insights, C., e micromobility revolution: How bikes and scooters are shaking up urban transport worldwide. 2019, Technical report, CB Insights.
15. Kegalle, H., et al., Are footpaths encroached by shared e-scooters? Spatio-temporal Analysis of Micro-mobility Services. arXiv preprint arXiv:2304.08721, 2023.
16. Glavić, D., et al., Influence of Dockless Shared E-Scooters on Urban Mobility: WTP and Modal Shift. Sustainability, 2023. 15(12): p. 9570.
17. Shah, N.R., et al., Why do people take e-scooter trips? Insights on temporal and spatial usage patterns of detailed trip data. Transportation Research Part A: Policy and Practice, 2023. 173: p. 103705.
18. Ravalet, E., Y. Dubois, and V. Kaufmann, Territories of High Mobility: Micro and Macro Analysis, in High Mobility in Europe: Work and Personal Life. 2015, Springer. p. 129-152.
19. Putra, I.M., A. Prakoso, and E.A. Setiawan. The utilization of e-scooter on green transit oriented development concept in tropical countries. in AIP Conference Proceedings. 2020. AIP Publishing.
20. Reis, A.F., P. Baptista, and F. Moura, How to promote the environmental sustainability of shared e-scooters: A life-cycle analysis based on a case study from Lisbon, Portugal. Journal of Urban Mobility, 2023. 3: p. 100044.
21. Javadiansr, M., et al., Coupling shared E-scooters and public transit: a spatial and temporal analysis. Transportation Letters, 2023: p. 1-18.
22. Pourfalatoun, S., J. Ahmed, and E.E. Miller, Shared Electric Scooter Users and Non-Users: Perceptions on Safety, Adoption and Risk. Sustainability, 2023. 15(11): p. 9045.
23. Ajzen, I., The theory of planned behavior. Organizational behavior and human decision processes, 1991. 50(2): p. 179-211.
24. Brown, S.A. and V. Venkatesh, Model of adoption of technology in households: A baseline model test and extension incorporating household life cycle. MIS quarterly, 2005: p. 399-426.
25. Handy, S., X. Cao, and P.L. Mokhtarian, Self-selection in the relationship between the built environment and walking: Empirical evidence from Northern California. Journal of the American planning association, 2006. 72(1): p. 55-74.
26. Broach, J., J. Dill, and J. Gliebe, Where do cyclists ride? A route choice model developed with revealed preference GPS data. Transportation Research Part A: Policy and Practice, 2012. 46(10): p. 1730-1740. [CrossRef]
27. Ewing, R. and R. Cervero, Travel and the built environment: A meta-analysis. Journal of the American planning association, 2010. 76(3): p. 265-294.
28. Tokey, A.I., S.A. Shioma, and S. Jamal, Analysis of spatiotemporal dynamics of e-scooter usage in Minneapolis: Effects of the built and social environment. Multimodal Transportation, 2022. 1(4): p. 100037.
29. McKenzie, G., Spatiotemporal comparative analysis of scooter-share and bike-share usage patterns in Washington, DC. Journal of transport geography, 2019. 78: p. 19-28.
30. Gelman, A. and J. Hill, Data analysis using regression and multilevel/hierarchical models. 2006: Cambridge university press.
31. Breiman, L., Random forests. Machine learning, 2001. 45: p. 5-32.
32. Almannaa, M.H., et al., Perception analysis of E-scooter riders and non-riders in Riyadh, Saudi Arabia: Survey outputs. Sustainability, 2021. 13(2): p. 863.
33. Button, K., H. Frye, and D. Reaves, Economic regulation and E-scooter networks in the USA. Research in transportation economics, 2020. 84: p. 100973.
34. Bai, S., et al., The relationship between E-scooter travels and daily leisure activities in Austin, Texas. Transportation research part D: transport and environment, 2021. 95: p. 102844.
35. Al-Thawadi, F.E. and S.G. Al-Ghamdi, Evaluation of sustainable urban mobility using comparative environmental life cycle assessment: A case study of Qatar. Transportation research interdisciplinary perspectives, 2019. 1: p. 100003.
36. Al-Mohannadi, M., et al., Sustainable Status Assessment of the Transit-Oriented Development in Doha’s Education City. Sustainability, 2023. 15(3): p. 1913.
37. Al-Thawadi, F.E., A.-A.A. Banawi, and S.G. Al-Ghamdi, Social impact assessment towards sustainable urban mobility in Qatar: Understanding behavioral change triggers. Transportation research interdisciplinary perspectives, 2021. 9: p. 100295.
38. Hardt, C. and K. Bogenberger, Usage of e-scooters in urban environments. Transportation research procedia, 2019. 37: p. 155-162.
39. Palantzas, G. and D. Nalmpantis, Data and perspectives on e-scooters use in Greece. Ekistics and The New Habitat, 2021. 81(1): p. 28-36.
40. Sellaouti, A., O. Arslan, and S. Hoffmann. Analysis of the use or non-use of e-scooters, their integration in the city of Munich (Germany) and their potential as an additional mobility system. in 2020 IEEE 23rd International Conference on Intelligent Transportation Systems (ITSC). 2020. IEEE.
41. Lee, M., et al., Forecasting e-scooter substitution of direct and access trips by mode and distance. Transportation research part D: transport and environment, 2021. 96: p. 102892.
42. Gössling, S., Integrating e-scooters in urban transportation: Problems, policies, and the prospect of system change. Transportation Research Part D: Transport and Environment, 2020. 79: p. 102230.
43. Bai, S. and J. Jiao, Dockless E-scooter usage patterns and urban built Environments: A comparison study of Austin, TX, and Minneapolis, MN. Travel behaviour and society, 2020. 20: p. 264-272.
44. Gebhart, K. and R.B. Noland, The impact of weather conditions on capital bikeshare trips. 2013.
45. Ashqar, H.I., M. Elhenawy, and H.A. Rakha, Modeling bike counts in a bike-sharing system considering the effect of weather conditions. Case studies on transport policy, 2019. 7(2): p. 261-268.
46. Scott, D.M. and C. Ciuro, What factors influence bike share ridership? An investigation of Hamilton, Ontario’s bike share hubs. Travel behaviour and society, 2019. 16: p. 50-58.
47. El-Assi, W., M. Salah Mahmoud, and K. Nurul Habib, Effects of built environment and weather on bike sharing demand: a station level analysis of commercial bike sharing in Toronto. Transportation, 2017. 44(3): p. 589-613.
48. Rejali, S., et al., Assessing a priori acceptance of shared dockless e-scooters in Iran. Transportation Research Part D: Transport and Environment, 2021. 100: p. 103042.
49. Abouelela, M., E. Chaniotakis, and C. Antoniou, Understanding the landscape of shared-e-scooters in North America; Spatiotemporal analysis and policy insights. Transportation research part A: policy and practice, 2023. 169: p. 103602.
50. Orvin, M.M., J.K. Bachhal, and M.R. Fatmi, Modeling the demand for shared e-scooter services. Transportation research record, 2022. 2676(3): p. 429-442.
51. Pobudzei, M., et al. Riders on the Storm: Exploring Meteorological and Temporal Impacts on Shared E-Scooters (SES) in Munich, Germany. in 2022 IEEE International Smart Cities Conference (ISC2). 2022. IEEE.
52. Tuli, F.M., S. Mitra, and M.B. Crews, Factors influencing the usage of shared E-scooters in Chicago. Transportation research part A: policy and practice, 2021. 154: p. 164-185.
53. Noland, R.B., Scootin’in the rain: Does weather affect micromobility? Transportation research part A: policy and practice, 2021. 149: p. 114-123.
54. Noland, R.B., Trip patterns and revenue of shared e-scooters in Louisville, Kentucky. Findings, 2019.
55. Kimpton, A., et al., E-scooters are becoming wildly popular–but we have to factor in the weather. 2022.
56. Kimpton, A., et al., Weather to scoot? How weather shapes shared e-scooter ridership patterns. Journal of Transport Geography, 2022. 104: p. 10343.
57. Corcoran, J., et al., Spatio-temporal patterns of a Public Bicycle Sharing Program: the effect of weather and calendar events. Journal of Transport Geography, 2014. 41: p. 292-305.
58. WeatherSpark. 2023.
59. Feng, Y., et al. Micromobility in smart cities: A closer look at shared dockless e-scooters via big social data. in ICC 2021-IEEE International Conference on Communications. 2021. IEEE.
60. Fearnley, N., E. Johnsson, and S.H. Berge, Patterns of e-scooter use in combination with public transport. Findings, 2020.
61. Oehme, R., et al., Data-based Usage Analysis of Shared e-scooters in the Context of Public Transport. 2022.
62. Mohammadian, A., et al., Analyzing the Impacts of a Successful Diffusion of Shared E-Scooters and Other Micromobility Devices and Efficient Management Strategies for Successful Operations in Illinois. FHWA-ICT-22-006, 2022.
63. Wang, K., et al., What travel modes do shared e-scooters displace? A review of recent research findings. Transport Reviews, 2023. 43(1): p. 5-31.
64. Kraussa, K., C. Dolla, and C. Thigpenb, The Net Sustainability Impact of Shared Micromobility in Six Global Cities. submitted to Case Studies on Transport Policy, 2022.
65. Mclean, R., C. Williamson, and L. Kattan. Simulation Modeling of Urban E-Scooter Mobility. in 2021 29th International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunication Systems (MASCOTS). 2021. IEEE.
66. Guiton, G. and F. Finlay, G174 (P) Electric scooters, their safety and the law around their use in the UK. 2020, BMJ Publishing Group Ltd.
67. Hosseinzadeh, A., et al., Spatial analysis of shared e-scooter trips. Journal of transport geography, 2021. 92: p. 103016.
68. Fil, M.M., Examining the Factors Leading to The use of E-Scooters in the Context of Micromobility. 2022, Marmara Universitesi (Turkey).
69. Alsulami, N., et al., Modeling Dockless Shared E-Scooter Demand by Time of Day: A Case Study of Austin. Journal of Advanced Transportation, 2023. 2023.
70. Caspi, O., M.J. Smart, and R.B. Noland, Spatial associations of dockless shared e-scooter usage. Transportation Research Part D: Transport and Environment, 2020. 86: p. 102396.
71. Jiao, J. and S. Bai, Understanding the shared e-scooter travels in Austin, TX. ISPRS International Journal of Geo-Information, 2020. 9(2): p. 135.
72. Cervero, R. and M. Duncan, Walking, bicycling, and urban landscapes: evidence from the San Francisco Bay Area. American journal of public health, 2003. 93(9): p. 1478-1483.
73. Ham, S.W., et al., Spatiotemporal demand prediction model for e-scooter sharing services with latent feature and deep learning. Transportation research record, 2021. 2675(11): p. 34-43.
74. Lin, L., Z. He, and S. Peeta, Predicting station-level hourly demand in a large-scale bike-sharing network: A graph convolutional neural network approach. Transportation Research Part C: Emerging Technologies, 2018. 97: p. 258-276.
75. Zhou, Y., et al., A Markov chain based demand prediction model for stations in bike sharing systems. Mathematical problems in engineering, 2018. 2018.
76. Kim, T.S., W.K. Lee, and S.Y. Sohn, Graph convolutional network approach applied to predict hourly bike-sharing demands considering spatial, temporal, and global effects. PloS one, 2019. 14(9): p. e0220782.
77. Kim, S., et al., Predicting Demand for Shared E-Scooter Using Community Structure and Deep Learning Method. Sustainability, 2022. 14(5): p. 2564.
78. Zhao, P., et al., A machine learning based approach for predicting usage efficiency of shared e-scooters using vehicle availability data. AGILE: GIScience Series, 2022. 3: p. 1-10.
79. Moosavi, S.M.H., et al., Understanding and Predicting the Usage of Shared Electric Scooter Services on University Campuses. Applied Sciences, 2022. 12(18): p. 9392.
80. Zhou, H., et al., Sustainable mobility strategies and their impact: a case study using a multimodal activity based model. Case Studies on Transport Policy, 2023. 11: p. 100945.
81. Uddin, M., H.-L. Hwang, and M.S. Hasnine, An interpretable machine learning framework to understand bikeshare demand before and during the COVID-19 pandemic in New York City. Transportation Planning and Technology, 2023: p. 1-17.
82. Zuniga-Garcia, N., et al., Evaluation of e-scooters as transit last-mile solution. Transportation research part C: emerging technologies, 2022. 139: p. 103660.
83. Dündar, S., E.Ş. Berktas, and M.C. Höke. Modelling the Effects of E-Scooters in Urban Traffic Using Artificial Neural Networks. in 2022 Innovations in Intelligent Systems and Applications Conference (ASYU). 2022. IEEE.
84. Elhenawy, M., et al., A novel crowdsourcing model for micro-mobility ride-sharing systems. Sensors, 2021. 21(14): p. 4636.
85. Yu, J., et al., Optimization model for the supply volume of bike-sharing: Case study in nanjing, china. Information, 2021. 12(5): p. 182.
86. Deng, J., et al., The pulse of urban transport: exploring the co-evolving pattern for spatio-temporal forecasting. ACM Transactions on Knowledge Discovery from Data (TKDD), 2021. 15(6): p. 1-25.
87. Yu, Z., et al., A Spatiotemporal Prediction Model for Regional Scheduling of Shared Bicycles Based on the INLA Method. Journal of Advanced Transportation, 2021. 2021: p. 1-10.
88. Ciociola, A., et al. E-scooter sharing: Leveraging open data for system design. in 2020 IEEE/ACM 24th International Symposium on Distributed Simulation and Real Time Applications (DS-RT). 2020. IEEE.
89. Wu, C.Y., C.S. Chiu, and P.F. Kuo. Defining the spatial impacts of changing the youbike fees: A Case Study of Taipei’s Elimination of the “first 30 minutes is free” Policy. in 40th Asian Conference on Remote Sensing: Progress of Remote Sensing Technology for Smart Future, ACRS 2019. 2020.
90. Wang, K. and Y.-J. Chen, Joint analysis of the impacts of built environment on bikeshare station capacity and trip attractions. Journal of Transport Geography, 2020. 82: p. 102603.
91. Zhang, X., et al. A dynamic shared bikes rebalancing method based on demand prediction. in 2019 IEEE Intelligent Transportation Systems Conference (ITSC). 2019. IEEE.
92. Guidon, S., et al., Electric bicycle-sharing: A new competitor in the urban transportation market? An empirical analysis of transaction data. Transportation research record, 2019. 2673(4): p. 15-26.
93. Zafar, U., et al. Analysis of GPS-based high resolution vehicle mobility data towards the electrification of transportation in Qatar. in IECON 2022–48th Annual Conference of the IEEE Industrial Electronics Society. 2022. IEEE.
94. Kåresdotter, E., et al., First mile/last mile problems in smart and sustainable cities: A case study in Stockholm County. Journal of Urban Technology, 2022. 29(2): p. 115-137.
95. Cheng, L., et al., Applying a random forest method approach to model travel mode choice behavior. Travel behaviour and society, 2019. 14: p. 1-10.
96. Bhaskar, A. and E. Chung, Passenger segmentation using smart card data. IEEE Transactions on intelligent transportation systems, 2014. 16(3): p. 1537-1548.
97. Lu, Y., et al., Commuting mode choice in a high-density city: Do land-use density and diversity matter in Hong Kong? International journal of environmental research and public health, 2018. 15(5): p. 920.
98. Breiman, L., Random forests. Machine learning, 2001. 45(1): p. 5-32.
99. Breiman, L., et al., Classification and regression trees. 1984.
100. Milakis, D., et al., Is micro-mobility sustainable? An overview of implications for accessibility, air pollution, safety, physical activity and subjective wellbeing. Handbook of sustainable transport, 2020: p. 180-189.