Traffic Flow Prediction Model Using Google Map and LSTM Deep Learning

Traffic jam is the most important factor of the urban road networks for making advanced travel plans, estimating traffic density and proactively managing the traffic flow. It causes adversely affect the social life, country economy, human health and is sometimes unable to manage traffic flow and signal. We explore the stacked long short-term memory (LSTM) network model to perform the multi-step ahead traffic speed prediction by employing Google Maps real-time and historical traffic data of three different types of urban road sections. After that, a Time-dependent correlation algorithm is used to map the predicted speed into the predicted traffic flow. The experimental results explored that, propose stacked LSTM model’s multi-step advanced predicted traffic flow mean relative error is varying between 8.25% ~14.09%. Also, results showed that the prediction accuracy improves and is stable with the freeway and identical traffic flow.