Predicting Traffic Volume and Occupancy at Failed Detectors

Accurate Traffic flow prediction relies on correctness of the values received from detectors. It is often the case that detectors are not working correctly and provide with incorrect values. The aim of this work is to predict the traffic flow variables at the failed detectors using deep learning techniques such as neural network and autoencoders. The major contributions are using neural network to model the complete network of detectors and use of autoencoders to reduce model size by exploiting spatial correlation between detectors. To the best of our knowledge deep learning has never been applied incase of detector failure.