A method for automatically learning parking preferences for a driver and generating parking recommendations includes generating training data based at least in part on: 1) trajectory data indicating a trajectory of a vehicle during a plurality of parking events, each in which a driver of the vehicle selected a parking candidate from among a plurality of parking candidates, and 2) attribute data indicating attributes of each of the plurality of parking candidates, removing, from the training data, data associated with at least one available parking candidate based on one or more conditions that indicate the driver did not consider the at least one available parking candidate, and training a decision model, based on remaining training data, to estimate a preferred parking candidate for the driver from among a set of available parking candidates.