A company collects sensor data (temperature, pressure, etc.) to predict equipment failures. Before training, the ML specialist must detect and remove outliers with minimal operational overhead. Which approach meets this requirement with the least effort?

The correct answer is to use Amazon SageMaker Data Wrangler's anomaly detection visualization and add a transformation to remove outliers. Data Wrangler is designed for data preparation, and its anomaly detection visualization provides a quick, low-code way to identify outliers. Once identified, you can directly add a transformation step within the same Data Wrangler flow to remove them, minimizing operational overhead. Incorrect options: Manually computing quartiles in a SageMaker Studio notebook and then using Data Wrangler is more labor-intensive and less integrated than using Data Wrangler's built-in visualization and transformation capabilities. A SageMaker Data Wrangler bias report focuses on identifying bias in the dataset, not specifically on detecting and removing general outliers for data cleaning. Amazon Lookout for Equipment is a fully managed service for anomaly detection in industrial equipment data, but it's designed for detecting anomalies in streaming data for predictive maintenance, not primarily for cleaning historical datasets by removing outliers before model training. While it detects anomalies, integrating its output back into a data cleaning workflow for pre-training data would involve more steps than using Data Wrangler directly.