Which solution will meet this requirement with the LEAST development effort?
Use Amazon Forecast with Holidays featurization and the built-in autoregressive integrated moving average (ARIMA) algorithm to train the model.
Use Amazon Forecast with Holidays featurization and the built-in DeepAR+ algorithm to train the model.
Use Amazon SageMaker Processing to enrich the data with holiday information. Train the model by using the SageMaker DeepAR built-in algorithm.
Use Amazon SageMaker Processing to enrich the data with holiday information. Train the model by using the Gluon Time Series (GluonTS) toolkit.
Explanations:
While Amazon Forecast with ARIMA could handle seasonal sales, it may not effectively leverage the correlations with other products and holidays without additional configuration. The ARIMA model typically requires more manual feature engineering and tuning compared to other models.
Amazon Forecast with Holidays featurization and the DeepAR+ algorithm is designed for time series forecasting with seasonality and can effectively utilize correlations with related products. It requires minimal development effort since it integrates these features out of the box.
Using SageMaker Processing to enrich the data with holiday information requires additional steps and configurations. While DeepAR is powerful, managing data preprocessing and training pipelines in SageMaker typically demands more development effort compared to using Amazon Forecast directly.
Similar to option C, using SageMaker Processing with GluonTS entails significant development effort for data enrichment and model training. GluonTS is a flexible toolkit but requires more coding and setup compared to using a fully managed service like Amazon Forecast.
From my perspective, the answer is:
Use Amazon Forecast with Holidays featurization and the built-in DeepAR+ algorithm to train the model.