Glosarium AI
Kamus lengkap Kecerdasan Buatan
MSE (Mean Squared Error)
Metric calculating the average of squared errors, penalizing large errors more than small errors in model evaluation.
RMSE (Root Mean Squared Error)
Square root of MSE, providing an error measure in the same unit as the target variable while maintaining the penalty for large errors.
R² (Coefficient of Determination)
Statistical indicator measuring the proportion of the target variable's variance explained by the model, ranging between 0 and 1 (or negative for very poor models).
MAPE (Mean Absolute Percentage Error)
Metric expressing the average error as a percentage of actual values, facilitating interpretation and comparison between models or datasets.
RMSLE (Root Mean Squared Logarithmic Error)
Variant of RMSE applied to the logarithms of values, particularly suitable for data with exponential distribution or to reduce the impact of outliers.
MedAE (Median Absolute Error)
Median of absolute errors, providing a robust measure against outliers unlike MAE which uses the mean.
Explained Variance Score
Metric evaluating the proportion of data variance explained by the model, similar to R² but without constraints on prediction bias.
Mean Squared Log Error (MSLE)
Metric calculating the average of squared logarithmic errors, ideal for predictions where relative errors are more important than absolute errors.
Quantile Loss
Asymmetric loss function used to predict specific quantiles of the conditional distribution, penalizing underestimations and overestimations differently.
Mean Absolute Scaled Error (MASE)
Relative metric comparing the mean absolute error of the model to that of a naive reference method, independent of data scale.
SMAPE (Symmetric Mean Absolute Percentage Error)
Symmetric variant of MAPE avoiding division by zero issues and providing better balance between overestimation and underestimation.
Mean Bias Error (MBE)
Metric measuring systematic bias of the model by calculating the mean of non-absolute errors, indicating whether the model tends to overestimate or underestimate.
Relative Absolute Error (RAE)
Normalized metric comparing the total absolute error of the model to that of a naive predictor, expressed as a unitless ratio.
Root Relative Squared Error (RRSE)
Square root of relative squared error, normalizing RMSE relative to the error of a simple reference model.
Theil's U Coefficient
Comparative forecasting metric measuring the relative performance of the model compared to a naive forecast, with U=0 indicating perfect prediction.
Nash-Sutcliffe Efficiency (NSE)
Efficiency coefficient ranging from -∞ to 1, measuring the model's ability to predict observed values relative to the mean of observations.
Mean Directional Accuracy (MDA)
Metric evaluating the model's ability to correctly predict the direction of changes between successive observations, crucial for financial applications.
Adjusted R²
Modified version of R² that adjusts the score based on the number of predictors in the model, avoiding artificial overvaluation when adding variables.