01698031

Component

A deep learning model is proposed 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 network (RNN) and 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, speed, transit, weather, etc. The model performance is evaluated on a case study conducted in Pittsburgh downtown area. Parking meter transactions, traffic speed, and weather data along with road networks are used in the case study. The proposed model outperforms other baseline methods including multi-layer LSTM and Lasso with a testing Mean Square Error (MSE) of 4.02 spaces when predicting block-level parking occupancies 30 minutes in advance. The case study also shows that features of traffic speed and weather are effective in predicting parking occupancies.

This paper was sponsored by TRB committee ABJ70 Standing Committee on Artificial Intelligence and Advanced Computing Applications.

19-05117

English

6p

2019

Transportation Research Board 98th Annual Meeting

Digital/other

Figures; References (11) ; Tables

Machine learning; Mathematical models; Mathematical prediction; Parking; Parking demand; Parking facilities; Parking turnover rate; Time periods; Traffic flow; Urban areas

Pittsburgh (Pennsylvania)

Data and Information Technology; Highways; Operations and Traffic Management

Transportation Research Board Annual Meeting 2019 Paper #19-05117

TRIS, TRB, ATRI

Dec 7 2018 9:44AM