Shenzhen Smart City Group rebuilt their traffic data platform on MatrixOne, reducing data components by 80%, achieving sub-second analytics, and cutting operational costs by 50%.
SUTPC (Shenzhen Urban Transport Planning Center) is the primary transportation big data and analytics organization for Shenzhen, one of China's largest and most congested megacities. It collects, integrates and analyzes traffic data from across the city's road network to support urban transport planning and real-time traffic management.
Predicting urban traffic congestion requires fusing dozens of heterogeneous data sources-road sensors, GPS trajectories, public transit ridership, weather feeds, event calendars, construction schedules-into unified datasets suitable for machine learning model training. Data engineers spend extensive time on ETL pipelines, feature engineering, and data quality management. Once ML models are trained and deployed, only specialized data scientists can operate them-traffic planners and city officials who need predictions and scenario analysis cannot interact with models directly, creating a bottleneck between insight generation and decision-making.
SUTPC deployed a Traffic Big Data + AI Prediction Platform on OmniFabric, combining multi-source data integration and ML model training for congestion prediction with an LLM-powered natural language interface that lets non-technical users query, configure, and interact with ML models conversationally.
Dozens of heterogeneous traffic data sources-sensors, GPS, transit, weather, events-are automatically integrated, cleaned, and transformed into ML-ready feature sets, dramatically reducing data engineering effort.
ML models trained on the unified dataset deliver congestion predictions at road-segment level, enabling proactive traffic management rather than reactive response.
The LLM-powered natural language interface lets traffic planners ask questions, run scenario simulations, and adjust model parameters in plain language-no coding or data science skills required.