Forecasting at Scale (Prophet)

Formulated automated business forecasting as a decomposable generalized additive model combining piecewise linear/logistic trends, Fourier seasonality, and holiday regressors.

The most widely deployed automated forecasting tool across tech, finance, and supply chain operations for business metrics and KPI dashboards.

Assumptions

• Business time series are dominated by macro human calendar effects (day-of-week, year-of-year) and trend shifts rather than high-order autoregressive noise

Limitations

• Ignores autoregressive error terms, yielding correlated residuals and suboptimal short-term 1-step-ahead forecasts