Exam AWS Certified Machine Learning - Specialty topic 1 question 216 discussion

Amazon Forecast is an AWS service that uses machine learning to build accurate time-series forecasts. It provides several built-in algorithms that support holiday featurization, and the DeepAR+ algorithm can handle the seasonality and correlation with other products with minimal development effort. With Amazon Forecast, the data scientist can easily configure the forecast horizon, select the appropriate forecast frequency, and configure the model training to incorporate the available historical data. Using Amazon SageMaker Processing to enrich the data with holiday information may require more development effort and does not offer the same level of automation and integration as Amazon Forecast.

While ARIMA is a classic time series forecasting method, it might not capture complex patterns (like seasonality and related product sales) as effectively as DeepAR+

B - DeepAR _ Algo with Holidya featurization in Amazon Forecast --> works better than Option A as Option A may be suboptimal but totally doable. DeepAR shines in multiple correlated time series and meant for seasonal data patterns

A. YES - fully managed solution B. NO - DeepAR+ is more for multiple time series ("In many applications, however, you have many similar time series across a set of cross-sectional units"- https://docs.aws.amazon.com/forecast/latest/dg/aws-forecast-recipe-deeparplus.html) C. NO - SageMaker DeepAR is too low-level D. NO - Gluon is too low-level

Option B is a great choice but requires more development effort over A which is also a great choice. Since the question asked for Least Development I am going with A

Deep AR can understand seasonal effect

B is correct

DeepAR accepts exogenorous regressors different from ARIMA and can understand seasonal effects, ARIMA can't do it too.

It is deepAR

Option B is a good choice, as the DeepAR+ algorithm is specifically designed for forecasting in time series data with seasonality and long-term dependencies. However, it may require more development effort compared to the ARIMA algorithm.

b https://docs.aws.amazon.com/forecast/latest/dg/holidays.html https://docs.aws.amazon.com/whitepapers/latest/time-series-forecasting-principles-with-amazon-forecast/appendix-a-faqs.html