Your team is developing a predictive maintenance solution for server failures using unlabeled monitoring data from a large data center. The solution must be scalable, cost-effective, and quickly implementable to prevent potential server downtimes. Given these constraints, what is the BEST first step to take? Choose one correct option.

Real Exam

Hire a team of qualified analysts to manually label the historical performance data, then train a model on this dataset.

12.0%

Create a simple heuristic, like a z-score, to label historical performance data and use this for real-time monitoring.

32.0%

Develop a heuristic to label historical data, then train a model to predict anomalies from this labeled dataset.

44.0%

Train a time-series model to predict performance values and alert on significant deviations from predictions.

12.0%

Google Professional Machine Learning Engineer