Traffic flow data with extreme events

This dataset provides processed urban traffic flow data with extreme-event annotations for four widely used open-source mobility datasets: NYC\u202fBike\u202f1 & 2, NYC\u202fTaxi, and BJ\u202fTaxi. Each dataset includes aggregated inflow\/outflow time series (1\u2011hour or 30\u2011minute intervals) across grid regions, chronologically split into train\/validation\/test sets (70\/10\/20\u202f%). We augment the original flows\u2014already publicly available\u2014with binary labels marking extreme events, defined as node-level traffic flows exceeding the 90th percentile. These annotations enable researchers to reproduce and extend our traffic forecasting approach, particularly for extreme events.