Smart Combined Forecast
Smart Combined Forecast is an ensemble method that combines precomputed forecasts from multiple methods by using a simple average. Smart Combined Forecast evaluates whether the two or three most accurate methods produce better results than the single most accurate method (the winner).
Ensemble forecasting combines predictions from multiple forecasting methods into a single forecast. Instead of selecting one method, such as ETS, ARIMA, or Prophet, the approach uses the strengths of several methods together.`
For example, you can collect advice from multiple experts and average the advice. The combined result can be better than advice from one expert.
Smart Combined Forecast evaluates whether a combined forecast from two or three methods produces a better result than the best-performing single method. If the combined forecast performs better, the Favorite view shows the result as Combined. Otherwise, the system uses the best-performing single method.
This feature works with DMP Competitive Forecasting and is metric-agnostic. The feature works with any provided error metric, such as MAE, MAPE, MSE, RMSE, MPE, ME, Theil’s U-Statistic, and Accuracy Signal. The method averages the provided errors and does not recalculate them. The feature also supports Symmetric Mean Absolute Percentage Error (SMAPE).
Key benefits
- Improved accuracy: Provides 5% to 20% higher accuracy than the best single method by combining complementary strengths, such as trend and seasonality handling.
- Reduced risk: Diversifies forecast risk across multiple methods. This approach makes the forecast less sensitive to individual method failures and more stable and reliable.
- Automatic optimization: Automatically evaluates which combination works best for the data. The method eliminates guesswork and adapts to specific demand patterns.