DataTerminology.comNEW
The number of future time steps a forecasting model predicts ahead.
DESCRIPTION
Forecast horizon specifies how far into the future predictions extend. It dictates model design, evaluation, and performance expectations. Short horizons often yield higher accuracy; longer horizons require robust temporal models and uncertainty quantification. Backtesting and rolling-origin evaluation depend on horizon choice. Business contexts define horizons via lead times and planning cycles. Communicating horizon is essential for stakeholders interpreting forecasts and risk.
ALTERNATIVE DEFINITIONS
Forecasting: Principles and Practice
The forecast horizon is the number of periods in the future for which forecasts are computed.
Clear textbook definition tied to periods ahead.
AWS Forecast
Forecast horizon is the number of time points that a predictor can forecast into the future.
Vendor framing applied to managed forecasting service.
Nixtla StatsForecast
h is the forecast horizon, i.e., the number of out-of-sample steps to predict.
Library convention clarifying horizon as out-of-sample steps.
RELATED TERMS
Time Series Forecasting Backtesting Prediction Interval Lead Time Rolling Forecast Temporal Aggregation ARIMA Cross-Validation Prophet
COMPETING TERMS
Nowcasting
KNOWLEDGE AREAS
Data Science Mathematics Data Visualization
TARGET AUDIENCE
Data Scientist BI Specialist Data Analyst Data Engineer
EXTERNAL SOURCES
- Wikipedia - Forecasting
- Hyndman & Athanasopoulos - Open textbook on forecasting
- statsmodels - Forecasting examples with horizons
- Prophet - Cross-validation with horizon setting
TOP QUESTIONS
How does Forecast Horizon affect model selection?
Longer horizons often need models capturing seasonality and trends robustly (e.g., ARIMA, ETS, ML ensembles), while short horizons can leverage simpler models.
Category: Requirements
What is the relationship between Forecast Horizon and accuracy?
Accuracy typically decreases as horizon increases due to compounding uncertainty; intervals should widen with longer horizons.
Category: Misunderstanding
How is Forecast Horizon used in backtesting?
Backtests simulate forecasts at each origin and evaluate errors at the specified horizon, often with rolling-origin evaluation.
Category: Testing
What business factors determine Forecast Horizon?
Lead times, planning cycles, shelf life, and demand volatility influence appropriate horizon choices.
Category: Management
Why communicate Forecast Horizon to stakeholders?
Horizon clarifies how far predictions extend, informing decisions, risk, and confidence in outcomes.
Category: Relevancy
Can different Forecast Horizons be combined?
Yes, multi-horizon models or hierarchical forecasting can produce short- and long-term forecasts simultaneously.
Category: Topic
How do uncertainty intervals change with Forecast Horizon?
Prediction intervals widen with horizon due to accumulating uncertainty, reflecting lower confidence farther out.
Category: Requirements
What are common mistakes with Forecast Horizon?
Using a horizon misaligned with decision timelines or evaluating models only at zero-step (nowcast) can mislead performance expectations.
Category: Mistakes
How to start learning about Forecast Horizon?
Study time series evaluation protocols in FPP3, experiment with rolling-origin backtests, and assess horizon-dependent metrics like MAPE or RMSE.
Category: Example