An ML engineer needs to use metrics to assess the quality of a time-series forecasting model. Which metrics apply to this model? (Choose two.)

Question

Options

• ARecall
• BLogLoss
• CRoot mean square error (RMSE)
• DInferenceLatency
• EAverage weighted quantile loss (wQL)

• A
  6% (1)
• B
  6% (1)
• C
  69% (11)
• D
  19% (3)

Explanation

RMSE (C) measures the average magnitude of prediction errors in the same units as the target variable, making it ideal for regression-style forecasting tasks where you care about how far off your predictions are. Average Weighted Quantile Loss (E) is specifically designed for probabilistic time-series forecasting, penalizing over- and under-predictions asymmetrically across quantiles - it's a core metric in services like Amazon Forecast.

Why the distractors are wrong:

• Recall (A) and LogLoss (B) are classification metrics - recall measures true positive rate, and log loss measures probabilistic classification confidence; neither applies to continuous value prediction.
• InferenceLatency (D) is an operational/infrastructure metric measuring how fast the model responds, not the quality of its forecasts.

Memory tip: For regression/forecasting tasks, think "How far off am I?" - that points to error-based metrics like RMSE. For probabilistic forecasts (predicting a range or distribution over time), think quantile loss (wQL). Any metric rooted in classification (recall, precision, F1, LogLoss) is a red flag when you see "time-series forecasting."

No community discussion yet for this question.