Modeling Sparse and Bursty Vulnerability Sightings: Forecasting Under Data Constraints

Researchers tested whether vulnerability sighting activity can be predicted over time.

They compared SARIMAX-style time-series models with count-based approaches such as Poisson regression.

The study found that sparse, bursty data makes standard time-series forecasting unreliable.

Count models and simpler decay methods were more stable for short-horizon prediction.

The results suggest cyber threat intelligence needs forecasting methods designed for rare, irregular events.