Abstract:The ability to predict city-wide parking availability is crucial for the successful development of Parking Guidance and Information (PGI) systems. Indeed, the effective prediction of city-wide parking availability can improve parking efficiency, help urban planning, and ultimately alleviate city congestion. However, it is a non-trivial task for predicting citywide parking availability because of three major challenges: 1) the non-Euclidean spatial autocorrelation among parking lots, 2) the dynamic temporal autocorrelation inside of and between parking lots, and 3) the scarcity of information about real-time parking availability obtained from real-time sensors (e.g., camera, ultrasonic sensor, and GPS). To this end, we propose Semi-supervised Hierarchical Recurrent Graph Neural Network (SHARE) for predicting city-wide parking availability. Specifically, we first propose a hierarchical graph convolution structure to model non-Euclidean spatial autocorrelation among parking lots. Along this line, a contextual graph convolution block and a soft clustering graph convolution block are respectively proposed to capture local and global spatial dependencies between parking lots. Additionally, we adopt a recurrent neural network to incorporate dynamic temporal dependencies of parking lots. Moreover, we propose a parking availability approximation module to estimate missing real-time parking availabilities from both spatial and temporal domain. Finally, experiments on two real-world datasets demonstrate the prediction performance of SHARE outperforms seven state-of-the-art baselines.

What problem does this paper attempt to address?

This paper attempts to solve the problem of predicting parking lot availability within the urban area. Specifically, the article proposes solutions to the following three main challenges: 1. **Non - Euclidean Spatial Autocorrelation**: The spatial associations between parking lots are not simple linear or Euclidean distance relationships, but complex and irregular. For example, the availability of a parking lot is not only affected by nearby parking lots, but may also change in synchronization with distant parking lots. 2. **Dynamic Temporal Autocorrelation**: The future availability of a parking lot is related to its availability in past time periods, and this temporal correlation changes over time. 3. **Scarcity of Real - time Parking Lot Availability Data**: Due to economic and privacy issues, only a small number of parking lots are equipped with real - time sensors (such as cameras, ultrasonic sensors, GPS, etc.), which leads to a shortage of real - time parking lot availability data. To address these challenges, the authors propose a Semi - supervised Hierarchical Recurrent Graph Neural Network (SHARE) model. The main contributions of this model include: - Proposing a semi - supervised spatio - temporal learning framework that combines environmental context factors and partially observed real - time parking lot availability data to predict parking lot availability within the urban area. - Introducing a hierarchical graph convolution module to capture the non - Euclidean spatial correlations between parking lots. This module consists of a Contextual Graph Convolution Block and a Soft Clustering Graph Convolution Block, which are used to model local and global spatial dependencies respectively. - Designing a parking lot availability approximation module to estimate the missing real - time availability of parking lots without sensor monitoring through the spatial and temporal domains. Specifically, a Propagating Convolution Block is introduced and the GRU module is reused in the temporal domain to approximate the missing parking lot availability, and then these two estimations are fused through an entropy - based mechanism. Through experiments on two real - world datasets (Beijing and Shenzhen), the results show that the prediction performance of the SHARE model is better than that of seven state - of - the - art baseline models. ### Summary of Involved Formulas 1. **Calculation Coefficient of Attention Mechanism**: \[ c_{ij} = \text{Attn}(W_a x_c^i, W_a x_c^j) \] where \( x_c^i \) and \( x_c^j \) are the current context representations of parking lots \( p_i \) and \( p_j \), \( W_a \) is the shared learning weight matrix, and \(\text{Attn}\) is the shared attention mechanism. 2. **Calculation of Proximity Score**: \[ \alpha_{ij} = \frac{\exp(c_{ij})}{\sum_{k \in N_i} \exp(c_{ik})} \] 3. **Update of Representation by Contextual Graph Convolution Operation**: \[ x_c'^i = \sigma\left( \sum_{j \in N_i} \alpha_{ij} W_c x_c^j \right) \] 4. **Calculation of Soft Assignment Matrix**: \[ S_{i,\cdot} = \text{Softmax}(W_s x_i) \] 5. **Generation of Latent Node Representation**: \[ x_s^i = \sum_{j = 1}^N S_{i,j}^\top x_j \] 6. **Soft Clustering Graph Convolution Operation**: \[ x_s'^i = \sigma\left( \sum_{j \in N_i} \alpha_s^{ij} W_l x_s^j \right) \] 7. **Gated Recurrent Unit (GR)

Semi-Supervised Hierarchical Recurrent Graph Neural Network for City-Wide Parking Availability Prediction

Residual spatial-temporal graph convolutional neural network for on-street parking availability prediction

Efficient Large-Scale Urban Parking Prediction: Graph Coarsening Based on Real-Time Parking Service Capability

Collaborative Parking Vacancy Prediction for Cities With Partial Sensors Missing

Spatio-temporal parking occupancy forecasting integrating parking sensing records and street-level images

Predicting vacant parking space availability zone-wisely: a graph based spatio-temporal prediction approach

Behavior-Aware Hypergraph Convolutional Network for Illegal Parking Prediction with Multi-Source Contextual Information

Predicting vacant parking space availability zone-wisely: a hybrid deep learning approach

A Short-Term Parking Demand Prediction Framework Integrating Overall and Internal Information

Hierarchical Graph Pooling is an Effective Citywide Traffic Condition Prediction Model

Dynamic Space Allocation Based on Internal Demand for Optimizing Release of Shared Parking

Parking Lot Traffic Prediction Based on Fusion of Multifaceted Spatio-Temporal Features

Intelligent Short-Term Multiscale Prediction of Parking Space Availability Using an Attention-Enhanced Temporal Convolutional Network.

Neural-Network-Based Dynamic Distribution Model of Parking Space Under Sharing and Non-Sharing Modes

Payment behavior prediction on shared parking lots with TR-GCN

Predicting Parking Availability in Singapore with Cross-Domain Data: A New Dataset and A Data-Driven Approach

Prediction of Parking Space Availability Using Improved MAT-LSTM Network

Beyond Prediction: On-street Parking Recommendation using Heterogeneous Graph-based List-wise Ranking

An Integrated Approach for the Near Real-Time Parking Occupancy Prediction

Leverage Multi-source Traffic Demand Data Fusion with Transformer Model for Urban Parking Prediction

Spatio-Temporal Graph Convolutional and Recurrent Networks for Citywide Passenger Demand Prediction