A deep learning approach to real-time parking occupancy prediction in transportation networks incorporating multiple spatio-temporal data sources

Abstract

A deep learning model is applied for predicting block-level parking occupancy in real time. The model leverages Graph-Convolutional Neural Networks (GCNN) to extract the spatial relations of traffic flow in large-scale networks, and utilizes Recurrent Neural Networks (RNN) with Long-Short Term Memory (LSTM) to capture the temporal features. In addition, the model is capable of taking multiple heterogeneously structured traffic data sources as input, such as parking meter transactions, traffic speed, and weather conditions. The model performance is evaluated through a case study in Pittsburgh downtown area. The proposed model outperforms other baseline methods including multi-layer LSTM and Lasso with an average testing MAPE of 10.6% when predicting block-level parking occupancies 30 minutes in advance. The case study also shows that, in generally, the prediction model works better for business areas than for recreational locations. We found that incorporating traffic speed and weather information can significantly improve the prediction performance. Weather data is particularly useful for improving predicting accuracy in recreational areas.

Short-Term Prediction of Demand for Ride-Hailing Services: A Deep Learning Approach

“…Those results show that the the use of spatio-temporal features can often yield superior predictive performance than RNN alone. Yang et al (2019) proposed a hybrid deep learning architecture, where the temporal features along with other traffic-related data are modelled using an LSTM model. The spatial correlations are modeled through a graph based convolutional neural network (GCN).…”

Spatiotemporal data mining: a survey on challenges and open problems

“…( 2015 ), IoT-Yang et al. ( 2019b ); Xue et al. ( 2016 ) Different learning models for varying spatiotemporal regions.…”

Hard to Park?

“…Yet another line of work employs techniques from image processing to determine the occupancy of parking spaces based on image data collected from various sources [5,9,12,16,17,31,34,35,39]. Yet other solutions have been proposed based on crowd-sourcing data from smartphones [40], GPS [26], and payments [32,41,42].…”

“…From a methodological perspective, various ML/statistical models for predicting parking occupancy have been proposed recently. These techniques cover a broad range of approaches including clustering [28,32,33], SVR [44], time-series analysis [22], Markov chains [21], vector autoregressive models [27], neural networks [1,19,29,36,41], and representation learning [45]. Particularly, Alajali et al [1] propose a Bayesian regularized neural network that takes into account historical data, traffic flow, and weather conditions.…”

“…Du-Parking [29] employs DNN and LSTM models to predict real-time parking availability, based on geolocation data from Baidu map and sensor data from parking lots. Yang et al [41] incorporate both CNN and RNN/LSTM for modeling spatiotemporal correlations, based on both real-time and historical data from multiple data sources including occupancy, conditions of traffic, road, and weather, as well as network topology. These learning approaches rely in part on parking lot sensor or parking meter data.…”