Skip to main content

Metrics

Evaluation metrics for quantifying model performance on regression and classification tasks.

Regression Metrics​

Mean Squared Error (MSE)​

The average of squared differences between predictions and true values. Penalizes large errors quadratically.

MSE(y,y^)=1n∑i=1n(yi−y^i)2\text{MSE}(y, \hat{y}) = \frac{1}{n} \sum_{i=1}^{n} (y_i - \hat{y}_i)^2

Root Mean Squared Error (RMSE)​

The square root of MSE, expressed in the same units as the target variable.

RMSE(y,y^)=MSE(y,y^)\text{RMSE}(y, \hat{y}) = \sqrt{\text{MSE}(y, \hat{y})}

Mean Absolute Error (MAE)​

The average of absolute differences. Less sensitive to outliers than MSE.

MAE(y,y^)=1n∑i=1n∣yi−y^i∣\text{MAE}(y, \hat{y}) = \frac{1}{n} \sum_{i=1}^{n} |y_i - \hat{y}_i|

Coefficient of Determination (R2R^2)​

Measures the proportion of variance in the target explained by the model. A score of 1.01.0 indicates perfect prediction; 0.00.0 corresponds to always predicting the mean yˉ\bar{y}.

R2(y,y^)=1−∑i=1n(yi−y^i)2∑i=1n(yi−yˉ)2R^2(y, \hat{y}) = 1 - \frac{\sum_{i=1}^{n} (y_i - \hat{y}_i)^2}{\sum_{i=1}^{n} (y_i - \bar{y})^2}

The score can be negative if the model is worse than predicting the mean.

#include <Skigen/Metrics>

Skigen::Metrics::mean_squared_error(y_true, y_pred);
Skigen::Metrics::root_mean_squared_error(y_true, y_pred);
Skigen::Metrics::mean_absolute_error(y_true, y_pred);
Skigen::Metrics::r2_score(y_true, y_pred);
FunctionDescription
mean_squared_errorMean of squared errors
root_mean_squared_errorSquare root of MSE
mean_absolute_errorMean of absolute errors
r2_scoreCoefficient of determination

Classification Metrics​

All binary classification metrics below are defined in terms of the confusion matrix entries:

  • TP (True Positives): correctly predicted positive
  • TN (True Negatives): correctly predicted negative
  • FP (False Positives): negative samples incorrectly predicted as positive
  • FN (False Negatives): positive samples incorrectly predicted as negative

Accuracy​

Accuracy=TP+TNn\text{Accuracy} = \frac{\text{TP} + \text{TN}}{n}

Precision​

Fraction of predicted positives that are truly positive. High precision means few false alarms.

Precision=TPTP+FP\text{Precision} = \frac{\text{TP}}{\text{TP} + \text{FP}}

Recall​

Fraction of actual positives that are correctly identified. High recall means few missed positives.

Recall=TPTP+FN\text{Recall} = \frac{\text{TP}}{\text{TP} + \text{FN}}

F1 Score​

The harmonic mean of precision and recall, balancing both concerns equally.

F1=2⋅Precision⋅RecallPrecision+Recall=2 TP2 TP+FP+FNF_1 = 2 \cdot \frac{\text{Precision} \cdot \text{Recall}}{\text{Precision} + \text{Recall}} = \frac{2\,\text{TP}}{2\,\text{TP} + \text{FP} + \text{FN}}
#include <Skigen/Metrics>

Skigen::Metrics::accuracy_score(y_true, y_pred);
Skigen::Metrics::precision_score(y_true, y_pred);
Skigen::Metrics::recall_score(y_true, y_pred);
Skigen::Metrics::f1_score(y_true, y_pred);
auto cm = Skigen::Metrics::confusion_matrix(y_true, y_pred);
FunctionDescription
accuracy_scoreFraction of correct predictions
precision_scorePrecision (positive predictive value)
recall_scoreRecall (sensitivity)
f1_scoreHarmonic mean of precision and recall
confusion_matrixN×N confusion matrix

Example​

#include <Skigen/LinearModel>
#include <Skigen/Metrics>
#include <Skigen/ModelSelection>

auto [X_train, X_test, y_train, y_test] = Skigen::train_test_split(X, y);

Skigen::LinearRegression model;
model.fit(X_train, y_train);
auto y_pred = model.predict(X_test);

std::cout << "MSE: " << Skigen::Metrics::mean_squared_error(y_test, y_pred) << "\n";
std::cout << "R²: " << Skigen::Metrics::r2_score(y_test, y_pred) << "\n";