Multiple horizon speed prediction for road networks

Intelligent Transport System (ITS) are advanced artificially intelligent systems that offer services pertaining to different traffic management and transport modes and enable the users to make a more informed and safer decision about the route they wish to take to reach their destination. Speed is a very important parameter when it comes to Intelligent Transport Systems. Predicting the speeds of vehicles at a future instant of time lets us know whether the traffic is going to be fairly smooth or congested. Also while obtaining data from loop detectors, sensors etc some of the values might be missing. Accurate prediction of the speed values can lead to the creation of low-dimensional models and also for missing data imputation. Loads of work has been done recently on predicting the speed values for a single link at a particular instant of time. However, our motivation was to predict speed values for multiple horizons simultaneously. Partial Least Squares (PLS), N-way PLS and Higher Order Partial Least Squares (HO-PLS) are the proposed models for this approach. 266 links were selected at random and the different prediction algorithms were trained. We were successful in predicting the speeds for 5 minutes, 1 0 minutes, 15 minutes and 30 minutes. N-way PLS slightly proved to be the best method for multiple horizon speed prediction for this particular dataset.Master of Science (Computer Control and Automation