Parking Availability Prediction with Long Short Term Memory Model

Traffic congestion causes heavily energy consumption, carbon dioxide emission and air pollution in cities, which is usually created by cars searching on-street parking spaces. Drivers are likely to move slowly and waste time on the road for an available on-street parking space if parking slot availability information is not revealed in advanced. Therefore, it is necessary for city councils to provide a car parking availability prediction service which could inform car drivers vacant parking slots before they start the journey. In this paper, we propose a novel framework based on recurrent network and use the long short-term memory (LSTM) model to predict parking multi-steps ahead. The core idea of this framework is that both the occupancy rate of on-street parking in a specific region and car leaving probability are exploited as prediction performance metric. A large real parking dataset is used to evaluate the proposed approach with extensive comparative experiments. Experimental results shows the proposed model outperform the state-of-art model.