All projects
system2025Smart CitiesPredictive ML
AION
Predictive traffic intelligence for emergency mobility
AIONVerified project view
0.971Test R²
44.5 veh/hTest MAE
17.4%Test sMAPE
01
Problem
Reactive navigation cannot anticipate a concert ending, announced roadworks or future city events that will change emergency accessibility.
02
Approach
A robust baseline, PCA embeddings, CatBoost and sigmoid shrink predict deviations from normality, while an LLM pipeline structures external events.
03
Scope
The public repository documents the capstone system and its roadmap; it does not present an operational emergency-service deployment.
Transparency
Limitations
Performance is weaker during low-signal early-morning hours, and API/cloud deployment remains on the roadmap.
Evidence
Sources
Next case