Answer: Symmetric Mean Absolute Percentage Error (sMAPE)

In Databricks AutoML, the default metric used to evaluate the performance of forecast models is Symmetric Mean Absolute Percentage Error (sMAPE). This metric is particularly suited for forecasting models as it considers both the magnitude and direction of errors, and is less sensitive to outliers compared to alternatives like MAPE. While MSE and RMSE are common for regression problems, they are not ideal for forecasting due to their heavy penalization of large errors. MAPE, though common in forecasting, can be skewed by forecasts close to zero, a limitation sMAPE addresses. Databricks AutoML likely defaults to sMAPE for its balanced and informative assessment of forecasting model performance. Note that depending on your specific needs, you may have the option to specify a different evaluation metric.

Author: LeetQuiz Editorial Team