Observational modeling uncovers population mobility patterns and event disruptions using mobile network data, indicating strong predictive utility for urban planning.
Key Points
To develop an end-to-end data pipeline for analyzing, detecting, and forecasting regional population mobility patterns and event-driven disruptions using cellular origin–destination records.
Analyzed 60 consecutive days of anonymized origin–destination mobile network data from 2019, covering millions of daily flows across census areas alongside demographic attributes.
Implemented a Z-score anomaly detection algorithm to identify unusual mobility disruptions triggered by localized events.
Trained a regional-scale regression forecasting model to predict incoming population flows and deployed the framework within a municipal decision-support tool.
The Z-score anomaly detection method effectively isolated and captured event-driven mobility surges and disruptions across the regional network.
The regression forecasting model attained an inbound-flow forecasting accuracy with a regional-scale mean absolute percentage error (MAPE) of approximately 10%.