Most route recommendation systems ignore the context of the driver (e.g. time of day) and the preferences of the driver (e.g. like to avoid highways). Not surprisingly, if a system can take this information into account, it can produce more appropriate routing directions. We apply novel machine learning techniques to learn driver preferences from actual driving data and use these to produce driving routes and predict driving destinations.

Link to Project
http://www.cs.cmu.edu/~pathway/