A company runs a three-tier web application in a VPC across multiple Availability Zones. Amazon EC2 instances in an Auto Scaling group host the application tier. The company needs an automated scaling plan that analyzes each resource's daily and weekly historical workload trends, and scales resources appropriately according to both forecasted and live utilization changes. Which scaling strategy should a solutions architect recommend?

The company requires scaling based on historical trends and forecasted demand, which is precisely what predictive scaling offers. Predictive scaling uses machine learning to forecast future traffic and proactively scales resources. Combining this with dynamic scaling using target tracking ensures that the Auto Scaling group maintains a specific utilization level (e.g., 70% CPU utilization) by automatically adjusting capacity based on live changes. Dynamic scaling with step scaling is reactive and doesn't account for historical trends or forecasts. Scheduled scaling is static and requires manual updates for changing patterns. Simple scaling policies are basic and lack the sophistication needed for trend analysis and forecasting.