Dense urban areas like Hamburg strive to increase the attractiveness of cycling to overcome mobility-related problems such as space limitations, air pollution, and noise levels. To address this issue, smart mobility solutions can encourage more people to choose cycling. Green Light Optimal Speed Advisory for bikes (bike-GLOSA) can reduce the number of stops at red lights, allow smoother traveling, and convey a digital advantage to cyclists. However, the city-wide implementation of bike-GLOSA introduces new challenges, including the need for automated lane matching. Humans may employ spatial reasoning to determine the most logical sequence of lanes and associated traffic lights across intersection topologies for a given route. In this paper, we aim to replicate this spatial reasoning using Geospatial Artificial Intelligence (GeoAI). The proposed Machine Learning (ML) model not only overcomes limitations associated with location- and camera-based lane matching approaches. It also outperforms a previous route-based approach by 8%, with an F1 score of 92% on our test dataset. We critically examine our approach and real-world results to identify potential limitations and avenues for future research.