Skip to content

Latest commit

 

History

History
20 lines (19 loc) · 2.05 KB

File metadata and controls

20 lines (19 loc) · 2.05 KB

Error Metrics

Indicator Name Key Characteristics
MAE Mean Absolute Error Measures average error magnitude with equal weighting, robust to outliers
RMSE Root Mean Squared Error Emphasizes larger errors through squaring, maintains original units
HUBER Huber Loss Combines MSE for small errors and MAE for large errors, balancing sensitivity and robustness
MSE Mean Squared Error Heavily weights large errors, produces squared units
ME Mean Error Preserves error direction to reveal systematic bias
MAPE Mean Absolute Percentage Error Expresses errors as percentages, enabling cross-scale comparison
SMAPE Symmetric Mean Absolute Percentage Error Improves on MAPE with symmetric treatment of errors
MPE Mean Percentage Error Reveals directional bias in percentage terms
MAPD Mean Absolute Percentage Difference Treats both series equally using average as denominator
MSLE Mean Squared Logarithmic Error Focuses on proportional rather than absolute errors
RMSLE Root Mean Squared Logarithmic Error Square root of MSLE, maintains better interpretability
MASE Mean Absolute Scaled Error Scales errors against naive forecast performance
RAE Relative Absolute Error Normalizes absolute errors against baseline model
RSE Relative Squared Error Normalizes squared errors against baseline model
RSQUARED R-Squared Measures proportion of variance explained by the model
DIRTY Dirty Data Injection Simulates missing data to test indicator robustness