Abstract
The rapid growth of modern megacities increases pressure on road networks and makes traditional public transport route planning methods insufficient.This paper presents a data driven approach for evaluating and optimizing urban route networks using mobile operator data, geospatial layers, and machine learning.The urban area is divided into a regular spatial grid, and a normalized load indicator based on unique subscribers is used as a proxy for spatiotemporal transport demand.This demand data is combined with information on stops, routes, and infrastructure to assess route performance.A gradient boosting model is trained to predict an integral route efficiency indicator reflecting the balance between passenger demand and vehicle capacity.The model is integrated into a web interface that enables route evaluation and automatic route generation.The methodology is transferable to other cities with available mobile and geospatial data and supports evidence based urban transport planning.