AI Forecasting Studio
Eight models compete on every series. The ensemble usually wins - here is where a simpler model is the smarter call, with the confidence to back it.
A real forecasting workbench — eight models compete and blend, with confidence intervals, accuracy metrics and AI-generated explanations executives can trust.
Best single model
Holt-Winters
3.3% wMAPE
Ensemble wMAPE
4.6%
-1.3%vs best single
Forecast confidence
93%
ensemble stability
Bias
+3.8%
near-neutral target
Forecast · Home Appliances ›
Actual history, model forecasts and 85% confidence band (ensemble)
37,119 u last actualDrill in →
Model comparison ›
Accuracy metrics across all eight methods (computed live)
| Model | wMAPE | RMSE | Bias | MAE | Confidence | Rank |
|---|---|---|---|---|---|---|
| Holt-Winters | 3.3% | 1,744 | +0.3% | 1,473 | Best | |
| Prophet-style | 3.5% | 1,930 | +0.1% | 1,530 | #2 | |
| Regression | 3.5% | 1,951 | +0.2% | 1,533 | #3 | |
| Ensemble | 4.6% | 2,319 | +3.8% | 1,876 | Production | |
| Weighted Moving Avg | 12.9% | 6,324 | +3.2% | 5,402 | #5 | |
| Moving Average | 13.4% | 6,499 | +2.7% | 5,647 | #6 | |
| Exponential Smoothing | 13.4% | 6,553 | +1.9% | 5,649 | #7 | |
| ARIMA / SARIMA | NaN% | NaN | NaN% | NaN | #8 |
AI forecast explanation
Plain-language rationale & confidence
For Home Appliances, the ensemble blends 7 models and lands at 4.6% wMAPE with 93% confidence and near-neutral bias (+3.8%). Holt-Winters is the strongest single model (3.3% wMAPE). This family shows strong seasonality and low volatility, so the model leans on Holt-Winters / SARIMA seasonal structure. The widening interval reflects growing uncertainty across the horizon.
Method library
Moving AverageWeighted Moving AvgExponential SmoothingHolt-WintersARIMA / SARIMAProphet-styleRegressionEnsemble