Data Fusion for Dynamic Traffic Prediction
Traffic congestion has huge negative impacts on the productivity, health and personal lives of city dwellers. To manage this problem effectively, transportation engineers need to predict traffic congestion throughout the road network at all hours of the day. Prediction of traffic typically involves travel surveys that are expensive, time consuming and do not capture temporal variation in travel demand. However, anonymised location data from mobile phones present an alternative source of data which is passively collected, widely available and naturally captures temporal trends. On the other hand, these data contain other biases and so if we use these data for transportation models, we must take care to correct for these biases using more reliable data. As part of the City Science collaboration with Andorra, we used a Bayesian network to build a calibrated transportation model for the country based on geolocated telecoms data and validated using a small sample of traffic counts.
Data Fusion for Dynamic Traffic Prediction
Traffic congestion has huge negative impacts on the productivity, health and personal lives of city dwellers. To manage this problem effectively, transportation engineers need to predict traffic congestion throughout the road network at all hours of the day. Prediction of traffic typically involves travel surveys that are expensive, time consuming and do not capture temporal variation in travel demand. However, anonymised location data from mobile phones present an alternative source of data which is passively collected, widely available and naturally captures temporal trends. On the other hand, these data contain other biases and so if we use these data for transportation models, we must take care to correct for these biases using more reliable data. As part of the City Science collaboration with Andorra, we used a Bayesian network to build a calibrated transportation model for the country based on geolocated telecoms data and validated using a small sample of traffic counts.