In recent years, to optimize transport systems and improve urban mobility, the analysis and comparison of urban transport routes has become crucial. This paper proposes a novel methodology for assessing the similarity of trajectories between key urban nodes, focusing on routes from an origin to a destination within a complex urban network. The approach aims to support tourists in the effective navigation of urban environments, and the approach is based on cosine similarity. The approach uses cosine similarity to quantify the similarity between travel trajectories based on node segment lengths. Applying this technique to a dataset of recorded urban trajectories provides valuable insights into mobility patterns, route efficiency and potential traffic bottlenecks. In addition, we are integrating these findings into a recommendation system based on machine learning. This system suggests optimal and alternative routes to users in real time. This system has important applications in transportation planning, intelligent traffic management, and emergency route optimization, and provides a data-driven approach to improving urban mobility. The results demonstrate the effectiveness of similarity-based trajectory analysis combined with machine learning in understanding urban movement dynamics and supporting decision-making in urban mobility planning.