A regression model must be especially sensitive to large prediction errors. Which metric should be evaluated alongside business requirements?

1. Classification precision
2. Training duration alone
3. The number of model parameters
4. Root mean squared error