Traffic Flow Forecasting at Micro-Locations in Urban Network using Bluetooth Detector

Predicting the urban traffic flow is of great importance for urban planners to be used in long-term prediction or in Intelligent Transport Systems (ITS) for short-term predictions. Traffic prediction is a challenging task because of complex spatial-temporal correlation between links in the road network. It is necessary to collect high-quality and rich-full traffic data for traffic state estimation and traffic prediction tasks. For this purpose, we investigate the ability of Bluetooth (BT) detector as a sensor at a micro-location to deliver additional information about the traffic condition. Furthermore, we used collected data to compare a few common time series methods: Random walk, Exponential smoothing, ARIMA, SARIMA, and Unobserved components. Our goal was to evaluate traffic data collected by a BT detector at a micro-location using time series forecasting methods. We showed that ARIMA model gives the best performance in forecasting a traffic demand. This data-driven approach can be helpful to inform drivers about better routing decisions and provides a guide for strategic traffic planning.