Multivariate Time Series Traffic Forecast with Long Short Term Memory based Deep Learning Model

The Intelligent Transportation System (ITS) is one of the key element to build smart cities. For the ITS traffic flow prediction plays a major role for a better traffic monitoring and control, route planning and for travel time estimation. With advent of sensor and Internet of Things (IoT) technologies a vast amount traffic data is now available which helps in performing the traffic prediction. Our proposed work is based on the Deep learning approach for short term traffic flow prediction. The correlation between various traffic parameters such as traffic flow, speed and occupancy are considered for the prediction. We performed the experiment with Recurrent Neural Network based multivariate Long Short Term Memory (LSTM) model. The results show the root mean square error (RMSE) of 9.8 and the mean absolute error (MAE) of 6.6. Thus when compared to the univariate approach the multivariate LSTM technique shows better performance and improves the accuracy of traffic flow prediction.