A healthcare provider wants to detect anomalous patient vital-sign patterns using an unlabeled dataset (health records, medication history, lifestyle). Which algorithm and hyperparameter choice is appropriate for unsupervised anomaly detection in SageMaker?

Random Cut Forest (RCF) is a suitable algorithm for unsupervised anomaly detection, especially with unlabeled data. RCF works by constructing a forest of random trees and identifying anomalies based on how easily data points are isolated. Setting numtrees to a value greater than 100 increases the robustness and accuracy of the anomaly detection process by creating a more diverse set of trees. XGBoost is a supervised learning algorithm for classification and regression, not unsupervised anomaly detection. K-means clustering can be used for anomaly detection by identifying outliers from clusters, but RCF is specifically designed for anomaly detection and often performs better in this context. DeepAR is a forecasting algorithm for time series data, not general anomaly detection.