You used 3 years of monthly inventory data for one SKU across 10 warehouses in Amazon Forecast. With DeepAR+ the predictor’s MAPE is much worse than the human forecasts. Which changes to the CreatePredictor API call could improve accuracy? (Choose two.)

Enabling PerformAutoML allows Amazon Forecast to automatically evaluate multiple predictor algorithms (like DeepAR+, Prophet, ARIMA, etc.) and select the best-performing one based on your dataset, which can significantly improve accuracy if DeepAR+ alone isn't optimal. Enabling PerformHPO (Hyperparameter Optimization) instructs Forecast to search for the best combination of hyperparameters for the chosen algorithm. This fine-tuning process can lead to a more accurate model by optimizing internal algorithm settings. Setting ForecastHorizon to 4 (e.g., 4 months) changes the prediction length but doesn't inherently improve model accuracy. Changing ForecastFrequency to 'W' (weekly) would require weekly data and might not be suitable for monthly input, potentially worsening results if not aligned with the data. Setting FeaturizationMethodName to 'filling' is a default for missing values and is unlikely to be the primary cause of poor performance compared to algorithm or hyperparameter choices, especially if the data is already clean.