Machine Learning Models for On-Street Parking Prediction: History
View Latest Version
Please note this is an old version of this entry, which may differ significantly from the current revision.
Contributor:
Saba Inam
,
Azhar Mahmood
,
Shaheen Khatoon
,
Majed Alshamari
,
Nazia Nawaz

Due to massive urbanization, traffic volume in urban areas has grown, making urban life very congested and polluted, leading to many negative impacts on human life, such as higher energy consumption, global warming, and airborne diseases. The goal of sustainable transport in smart cities is to ensure efficient traffic movement while minimizing a negative impact on the environment and public health.

  • smart city applications
  • Internet of Things
  • predictive analytics

1. Introduction

Due to massive urbanization, traffic volume in urban areas has grown, making urban life very congested and polluted, leading to many negative impacts on human life, such as higher energy consumption, global warming, and airborne diseases [1]. According to World Resource Institute [2], 74% of CO2 is produced by greenhouse gas emissions, and 93% of it results from fossil fuel usage, transportation, manufacturing, and consumption. In fact, 2020 has been recorded as the hottest year per NASA analysis. For the sustainable development of cities, the efficient use of resources and the adoption of effective measures have become crucial for survival. The researchers have witnessed the COVID-19 effects in different areas of life, making the internet and information the heart of modern and sustainable cities. To reap the benefits of the internet and Information Communication Technologies (ICT), many city governments have initiated the concept of a “Smart City” with the deployment of advanced ICT aiming to provide a better living experience to its citizens [3]. At the heart of a smart city is the Internet of Things (IoT) which enables different devices to interact and draws upon various underlying operations of a smart city for sustainable living such as smart services, smart health, smart transportation, smart agriculture, smart energy to name a few [4].
The goal of sustainable transport in smart cities is to ensure efficient traffic movement while minimizing a negative impact on the environment and public health [5]. The most discussed area in smart cities is intelligent transportation highlighting its impact on intelligent mobility, the environment, and the economy. For example, Cisco Barcelona Jurisdiction Profile 2014 [6] reveals an annual increase of $50 million through parking fee revenues using smart parking technology. The main goal of smart parking is finding and providing appropriate parking for each user. However, the problem of finding a parking area is still challenging due to increased traffic flow in urban areas. For example, some studies reveal that an average of 30–40% increase in traffic is caused by drivers looking for vacant parking spots, and on average, a New York driver spends 107 h a year searching for a parking spot [7]. This phenomenon has increased air pollution and has had a negative environmental effect. With an efficient parking infrastructure, the cities can reduce carbon emissions caused by additional fuel combustion and avoid delays and traffic congestion while looking for a free parking slot.
In previous research, smart parking solutions are mainly categorized as off-street and on-street [8][9]. Off-street parking includes garages and closed parking spaces which could be outdoors or indoors. Off-street, the problem is simpler since it is straightforward to count the number of available slots by counting the number of cars entering and leaving a closed parking space. Off-street parking management has been tackled quite well due to its simpler problem and data availability [10]. On the other hand, on-street parking is challenging due to the absence of parking entrances and significant changes in occupancy rates as more cars enter and leave the spots. On-street parking can directly affect streets regarding traffic congestion and air pollution. Numerous research has been done on both problems leading to an effective search for vacant parking spots. The research is usually based on parking spaces equipped with sensors to sense whether the spots are occupied and provide information. The data from occupancy sensors allows to learn availability patterns and predict probabilities of parking occupancy of the spots. Based on Parking Sensor Data (PSD), various machine learning methods have been used to predict parking occupancy rates [1][11]. The most common ML used for parking prediction are Regression Trees [12][13][14], DTs [14], Support Vector Machine [13][15], Genetic Algorithm [16], Bayesian [17], and Neural Network [13][18]. The performance of these models depends on the accuracy of information provided to users about the availability of parking lots. However, multiple traffic factors may influence car parking activity regarding on-street parking. For example, the occupancy status may change due to other traffic factors present at that time, such as weather, pedestrian mobility, and traffic volume; therefore, the information provided by PSD is not very efficient. These factors can influence car parking conditions; therefore, it is essential to identify possible factors to predict future parking availability accurately.
It is necessary to install many sensors in various cities with substantial setup costs to collect data for contextual factors. Using a publicly available dataset can provide a good starting point for understanding the impact of external elements on the real-time prediction of the availability of parking spaces.

2. Machine Learning Models for On-Street Parking Prediction

The prediction of car park availability is the subject that has received significant attention in the context of smart cities where parking facilities have installed sensors as part of their infrastructure. Many research efforts have focused on improving parking search efficiency, reservation, and prediction for an available parking space. For example, Kizilkaya et al. [19] used a hierarchical approach for predicting free parking spots using a binary search tree (BST). For the experiment, synthetic data is used with attributes such as parking distance, capacity, and availability status. The approach first searches for the nearest parking location and then finds a free spot in the nearest car park. Horng [20] used the Artificial Fish Swarm Algorithm(AFSA) to minimize search time and traffic congestion. The performance is evaluated through simulations by randomly distributing 300–1800 vehicles in a 5.0 × 5.0 km2 field. The results are compared with the conventional opportunistic methods revealing the effectiveness of AFSA in terms of reducing search time and congestion. However, the studies discussed above are promising, but the focus is limited to exploring only algorithmic capability in the domain.
Similarly, Thomas and Kovoor [16] used Genetic Algorithm (GA) to solve the scheduling problem in the parking system, but the proposed prototype can only be used for reserving parking spots. The decision-making of parking slots is based on the maximized fitness score of the GA objective function. The performance analysis metrics include efficiency, utilization, and average waiting time. Customers can book a parking slot in advance for a specific time period. When parking time duration exceeds, the system sends a notification of time exceeded. All the parking information is stored in the cloud. A multi-criteria decision analysis-based Parking space Reservation (MCPR) algorithm is proposed by Rehena et al. [21] for improvement in the reservation algorithm. The MCPR automatically finds the nearest parking space based on the users’ preferences, from parking space availability to pricing for the reservations. The studies mentioned in this section is a step further in optimizing the decision capability for predicting parking slot and highlighting the significance of using Machine Learning (ML) algorithms.
The previous research efforts indicate that several ML algorithms have been widely studied and explored in the direction of predicting occupancy. For example, Raj et al. [22] used parking data and tested the Random Forest method for predicting parking spots in a parking lot. The question is how contextual data and other ML methods can predict parking availability. Stolfi et al. [23] used historical car parking occupancy data from the Birmingham city council for testing various prediction strategies such as polynomial fitting, Fourier series, K-means clustering, and time series to predict future occupancy. The results are validated using K- fold cross-validation with the final output testing on unseen occupancy data. The solution is made available for the users through a webpage. However, it faces challenges due to the inconsistency in the sensor’s data, as the data may not be updated for the whole day. Klandev et al. [24] used garage occupancy and traffic congestion data to predict the parking spot availability ratio within 60 min. They tested the XGBoost regression model, which received a low error rate confirming its efficiency of predictions.
Similarly, Claudio et al. [17] compared different prediction techniques utilizing traffic flow, weather, and historical data to predict parking in the city garages of Florence. The resulting solution proved the Bayesian regularized network for reliable and fast predictions. Zheng et al. [13] perform a comparative analysis of SVM, Regression Tree, and Neural Networks using San Francisco and Melbourne datasets to predict long-term occupancy in 24 h intervals. The results indicate that the Regression tree outperforms the other two methods they evaluated with the highest accuracy and minimum error rate.
The discussed research shows the importance of contextual data such as traffic and weather being used along with ML approaches, but they are only tested on off-street parking prediction. Alajali et al. [12] investigated the use of on-street car parking, pedestrian, and daily traffic count data to predict short-term parking slots using Boosting Regression Tree. The research was implemented for a particular location Central Business District Melbourne, using data for special days and events since getting pedestrian counts were costly and hard to scale. Here, only in one research, the impact of pedestrian data and traffic is utilized for on-street parking. The results show that multisource data had an improved performance using gradient boosting (GBRT) with MSE 0.029. Still, the results are reported with only pedestrian data as the traffic data lacked proper mapping with other sources due to limited availability.
One of the critical challenges in addressing parking prediction is considering the nature of underlying data, suitable predictive models, and the accuracy of real-time decision-making. All the research done has focused on either one or the other challenge. For example, Liu et al. [8] proposed an online parking guidance system considering the delay in real-time parking space availability. The researchers discussed the multiuser online street problem, and the research was validated on a Melbourne dataset. The results illustrated that the proposed framework reduces 63.8% delay.
On the other hand, Vlahogianni et al. [25] proposed a two-step methodological framework for real-time car occupancy prediction based on sensor data. The first step predicts the real-time parking space using Recurrent Neural Networks (RNN). The second module is based on finding the available parking space with traffic volume. This approach, however, proved computationally expensive.
Among the studies discussed above, each has tried to solve different problems, such as minimizing delay and congestion, techniques to deal with inconsistent sensor data, and ML methods to improve the prediction accuracy for on-street and off-street parking prediction. Many studies used only car parking sensor data to evaluate the predictive performance of ML methods. Only a few studies focus on contextual factors such as traffic flow, weather, or pedestrian mobility data. However, these studies are evaluated on off-street prediction problems such as city garages and parking lots, where data accessibility of occupancy status is easier to obtain. None of the existing studies has investigated the relationship of car occupancy data with weather conditions, traffic count, and pedestrian mobility in their predictive models for on-street parking.
Additionally, most studies are evaluated via simulation, and very few are evaluated in real-time. Real-time studies lack computational scalability, which is crucial for today’s smart city applications. There have also been gaps in one way or the other, such as taking advantage of multisource data for on-street parking and making a solution scalable for real-time predictions. This work proposes a scalable predictive solution for real-time on-street parking prediction utilizing multisource data.

This entry is adapted from the peer-reviewed paper 10.3390/su14127317

References

  1. Abdeen, M.; Nemer, I.; Sheltami, T. A Balanced Algorithm for In-City Parking Allocation: A Case Study of Al Madinah City. Sensors 2021, 21, 3148.
  2. 4 Charts Explain Greenhouse Gas Emissions by Countries and Sectors. Available online: https://www.wri.org/insights/4-charts-explain-greenhouse-gas-emissions-countries-and-sectors (accessed on 13 March 2022).
  3. Wong, M.S.; Wang, T.; Ho, H.C.; Kwok, C.Y.T.; Lu, K.; Abbas, S. Towards a Smart City: Development and Application of an Improved Integrated Environmental Monitoring System. Sustainability 2018, 10, 623.
  4. Syed, A.; Sierra-Sosa, D.; Kumar, A.; Elmaghraby, A. IoT in Smart Cities: A Survey of Technologies, Practices and Challenges. Smart Cities 2021, 4, 429–475.
  5. Kurek, A.; Macioszek, E. Impact of Parking Maneuvers on the Capacity of the Inlets of Intersections with Traffic Lights for Road Traffic Conditions in Poland. Sustainability 2021, 14, 432.
  6. CISCO. IoE-Driven Smart City Barcelona Initiative Cuts Water Bills, Boosts Parking Revenues, Creates Jobs & More; 2014. Available online: https://www.cisco.com/c/dam/m/en_us/ioe/public_sector/pdfs/jurisdictions/Barcelona_Jurisdiction_Profile_final.pdf (accessed on 3 March 2020).
  7. Fadi, A.-T.; Arman, M. Smart parking in IoT-enabled cities: A survey. Sustain. Cities Soc. 2019, 49, 101608.
  8. Liu, K.S.; Gao, J.; Wu, X.; Lin, S. On-Street Parking Guidance with Real-Time Sensing Data for Smart Cities. In Proceedings of the 15th Annual IEEE International Conference on Sensing, Communication, and Networking (SECON), Hong Kong, China, 11–13 June 2018.
  9. Rajabioun, T.; Ioannou, P.A. On-Street and Off-Street Parking Availability Prediction Using Multivariate Spatiotemporal Models. IEEE Trans. Intell. Transp. Syst. 2015, 16, 2913–2924.
  10. Monteiro, F.V.; Ioannou, P. On-Street Parking Prediction Using Real-Time Data. In Proceedings of the 21st IEEE International Conference on Intelligent Transportation Systems (ITSC), Maui, HI, USA, 4–7 November 2018.
  11. Awan, F.M.; Saleem, Y.; Minerva, R.; Crespi, N. A Comparative Analysis of Machine/Deep Learning Models for Parking Space Availability Prediction. Sensors 2020, 20, 322.
  12. Alajali, W.; Wen, S.; Zhou, W. On-Street Car Parking Prediction in Smart City: A Multi-source Data Analysis in Sensor-Cloud Environment. In Proceedings of the International Conference on Security, Privacy and Anonymity in Computation, Communication and Storage, Guangzhou, China, 12–15 December 2017.
  13. Zheng, Y.; Rajasegarar, S.; Leckie, C. Parking availability prediction for sensor-enabled car parks in smart cities. In Proceedings of the 2015 IEEE Tenth International Conference on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP), Singapore, 7–9 April 2015.
  14. Jelen, G.; Podobnik, V.; Babic, J. Contextual prediction of parking spot availability: A step towards sustainable parking. J. Clean. Prod. 2021, 312, 127684.
  15. Matijosaitiene, I.; McDowald, A.; Juneja, V. Predicting Safe Parking Spaces: A Machine Learning Approach to Geospatial Urban and Crime Data. Sustainability 2019, 11, 2848.
  16. Thomas, D.; Kovoor, B.C. A Genetic Algorithm Approach to Autonomous Smart Vehicle Parking system. Procedia Comput. Sci. 2018, 125, 68–76.
  17. Claudio, B.; Paolo, N.; Irene, P. Predicting available parking slots on critical and regular services by exploiting a range of open data. IEEE Access 2018, 6, 44059–44071.
  18. Camero, A.; Toutouh, J.; Stolfi, D.H.; Alba, E. Evolutionary Deep Learning for Car Park Occupancy Prediction in Smart Cities. In Proceedings of the International Conference on Learning and Intelligent Optimization, Kalamata, Greece, 10–15 June 2018.
  19. Kizilkaya, B.; Caglar, M.; Al-Turjman, F.; Ever, E. Binary search tree based hierarchical placement algorithm for IoT based smart parking applications. Internet Things 2018, 5, 71–83.
  20. Horng, G.-J. The Adaptive Recommendation Mechanism for Distributed Parking Service in Smart City. Wirel. Pers. Commun. 2014, 80, 395–413.
  21. Rehena, Z.; Mondal, M.A.; Janssen, M. A Multiple-Criteria Algorithm for Smart Parking: Making Fair and Preferred Parking Reservations in Smart Cities. In Proceedings of the 19th Annual International Conference on Digital Government Research: Governance in the Data Age, Delft, The Netherlands, 30 May–1 June 2018.
  22. Raj, S.U.; Manikanta, M.V.; Harsitha, P.S.S.; Leo, M.J. Vacant Parking Lot Detection System Using Random Forest Classification. In Proceedings of the 3rd International Conference on Computing Methodologies and Communication (ICCMC), Erode, India, 27–29 March 2019.
  23. Stolfi, D.H.; Alba, E.; Yao, X. Predicting Car Park Occupancy Rates in Smart Cities. In Proceedings of the International Conference on Smart Cities, Malaga, Spain, 14–16 June 2017; Springer: Berlin/Heidelberg, Germany, 2017.
  24. Klandev, I.; Tolevska, M.; Mishev, K.; Trajanov, D. Parking Availability Prediction Using Traffic Data Services. In Proceedings of the IEEE Proceedings of the ICT Innovations, Online, 24–26 September 2020.
  25. Vlahogianni, E.I.; Kepaptsoglou, K.; Tsetsos, V.; Karlaftis, M.G. A Real-Time Parking Prediction System for Smart Cities. J. Intell. Transp. Syst. 2015, 20, 192–204.
More
© Text is available under the terms and conditions of the Creative Commons Attribution (CC BY) license; additional terms may apply. By using this site, you agree to the Terms and Conditions and Privacy Policy.
This entry is offline, you can click here to edit this entry!
Video Production Service
Feedback