White paper · February 2026
Minimizing Spatiotemporal Mismatch in Urban Logistics: A Topology-Aware Poisson Boosting Framework (Hex-PGBM)
Supply (drivers) and demand (riders) in urban mobility are rarely aligned, and human dispatch relies on intuition that produces empty kilometres and lost revenue. The paper presents the Bugama Predictive Oracle: a forecasting engine on level-8 H3 hexagons with a Poisson-objective gradient boosting model (Hex-PGBM), fed with weather, calendar cycles and event gravity models, plus a new Adaptive Topological Smoothing step for sparse zones.
18.4%lower error than baseline models15%projected gain in fleet utilisation