Model error analysis provides insight into how model performance changes over time on normalized data. This insight allows us to monitor whether the model is consistent and stable in its predictions. Specifically, we look at two indicators.
Mean absolute error (MAE)
This metric shows how close the model predictions are to the actual values. A lower MAE value means the model makes smaller errors, indicating higher accuracy of the predictions. Simply put, MAE is the average magnitude of the model’s error.
R² (R-squared) score
This metric shows how well the model explains the variability in the actual values. An R² value close to 1 indicates that the model explains changes in values very well. Thus, the R² score helps us understand how effectively the model captures patterns in the data.
If the MAE and R² values are stable over time, it means that the model is robust and its predictions are consistent.