You must maximize AUC for a binary classification problem using XGBoost and need to search for good values of eta, alpha, min_child_weight, and max_depth with minimal operational overhead. Which approach requires the least operational effort?

Amazon SageMaker automatic model tuning (Hyperparameter Optimization or HPO) is designed precisely for this scenario. It automates the process of finding the best hyperparameter values for a given algorithm and objective metric (like AUC) by intelligently searching a defined range of values. This significantly reduces operational effort compared to manual tuning or setting up custom distributed environments. Incorrect options: Bootstrap scikit-learn on an EMR cluster and run k-fold cross-validation on the algorithm: While technically feasible, this involves significant operational overhead in setting up, configuring, and managing an EMR cluster, installing libraries, and orchestrating the cross-validation process manually. Use prebuilt SageMaker Docker images that include scikit-learn and run k-fold cross-validation: This is better than EMR but still requires manual scripting and orchestration of the k-fold cross-validation and hyperparameter search, which is exactly what SageMaker HPO automates. Purchase an AUC-focused algorithm from AWS Marketplace and provide ranges for each hyperparameter: AWS Marketplace offers pre-trained models or algorithms, but it doesn't inherently provide an automated hyperparameter tuning service like SageMaker HPO. You would still need a mechanism to perform the tuning.