This bachelor thesis provides an implementation that solves the problem of learning individual route preferences of drivers, focusing on cycling routes and metrics for cyclists. Currently, many routing services consider only distance or travel time when calculating a route between some source and target, but there may be many other criteria that a user has in mind when planning a route, such as height ascent, route landscape, noise exposure along the route, suitability for cyclists and many more. Specifying each of these criteria is a non-trivial, unintuitive and time-consuming task for a user. The aim of this bachelor thesis is therefore to take the task of specifying such a preference out of the user’s responsibility by automating the process. We present an algorithm that deduces a user preference from a given path and splits the path if no such preference can be found for the entire length. In addition, we provide an implementation utilizing this algorithm, which features an intuitive, state-of-the-art front end that allows a cyclist to specify his personal routes on a map, and a back end that calculates the preferences and handles user administration.