What is the name of the supervised learning algorithm that forecasts scalar time series using recurrent neural networks (RNN)?

The algorithm used for forecasting scalar time series using recurrent neural networks (RNN) is known as Amazon SageMaker DeepAR. This algorithm is specifically designed to model and predict time-dependent data, making it well-suited for scenarios where forecasting future values in a sequential dataset is required.

DeepAR utilizes RNN architecture to capture temporal dependencies in the data, allowing it to learn patterns and trends over time. It can handle complex, non-linear relationships in time series data, and it excels in situations where multiple time series share similar characteristics, leveraging the correlated information across them for improved predictions.

In contrast, the other options serve different purposes: Amazon SageMaker XGBoost is primarily a decision tree-based ensemble algorithm for regression and classification tasks, while Amazon SageMaker Linear Learner focuses on linear regression and classification problems. Random Cut Forest is used for anomaly detection rather than forecasting. Each of these algorithms has specific applications that differ from the time series forecasting capabilities provided by DeepAR.