Documentation/Nodes/Evaluation Metrics

Evaluation Metrics

Comprehensive model evaluation toolkit with 3 specialized nodes for assessing classification and regression model performance using industry-standard metrics.

Node Reference

Detailed documentation for each evaluation node available in VERA.

Confusion Matrix

Create confusion matrix for classification model evaluation

Type: eval_confusion_matrixCategory: Classification Metrics

Key Features

  • •Multi-class support
  • •Normalized matrices
  • •Visual heatmap representation
  • •Class-wise statistics
  • •Error analysis capabilities

Input Ports

y_truedata

True target values

y_preddata

Predicted target values

labelsdata

Class labels (optional)

Output Ports

confusion_matrixdata

Confusion matrix array

matrix_plotimage

Visual confusion matrix

classification_metricsdata

Basic metrics from matrix

Classification Report

Generate detailed classification performance report

Type: eval_classification_reportCategory: Classification Metrics

Key Features

  • •Precision, recall, F1-score
  • •Support counts
  • •Macro/micro averaging
  • •Per-class metrics
  • •Overall accuracy

Input Ports

y_truedata

True target values

y_preddata

Predicted target values

target_namesdata

Class names (optional)

Output Ports

classification_reportdata

Detailed metrics report

report_plotimage

Visual report representation

macro_averaged_metricsdata

Overall performance metrics

Regression Metrics

Calculate regression performance metrics

Type: eval_regression_metricsCategory: Regression Metrics

Key Features

  • •Mean Absolute Error (MAE)
  • •Mean Squared Error (MSE)
  • •Root Mean Squared Error (RMSE)
  • •R-squared (coefficient of determination)
  • •Mean Absolute Percentage Error (MAPE)

Input Ports

y_truedata

True target values

y_preddata

Predicted target values

Output Ports

regression_metricsdata

Comprehensive metrics report

residual_plotimage

Residuals vs predictions plot

metrics_summarydata

Key metrics summary

Workflow Examples

Common model evaluation workflows using these specialized metric nodes.

Classification Model Evaluation

Complete evaluation pipeline for classification models

  1. 1Split data into train/test sets
  2. 2Train classification model
  3. 3Make predictions on test set
  4. 4Generate confusion matrix
  5. 5Create detailed classification report
  6. 6Analyze per-class performance
  7. 7Identify misclassification patterns

Regression Model Assessment

Comprehensive evaluation of regression model performance

  1. 1Prepare regression dataset
  2. 2Train regression model
  3. 3Generate predictions
  4. 4Calculate regression metrics
  5. 5Analyze residuals distribution
  6. 6Check for heteroscedasticity
  7. 7Validate model assumptions

Model Comparison Study

Compare multiple models using standardized metrics

  1. 1Train multiple models
  2. 2Evaluate each with confusion matrix
  3. 3Generate classification reports
  4. 4Calculate regression metrics
  5. 5Compare performance across models
  6. 6Select best performing model
  7. 7Document evaluation results

Evaluation Metric Categories

Comprehensive evaluation framework covering all major machine learning model types.

Classification Metrics

Basic Metrics

  • • Accuracy: (TP + TN) / (TP + TN + FP + FN)
  • • Precision: TP / (TP + FP)
  • • Recall: TP / (TP + FN)
  • • F1-Score: 2 * (Precision * Recall) / (Precision + Recall)

Advanced Metrics

  • • Specificity: TN / (TN + FP)
  • • NPV: TN / (TN + FN)
  • • MCC: Matthews Correlation Coefficient
  • • Cohen's Kappa

Regression Metrics

Error Metrics

  • • MAE: Mean Absolute Error
  • • MSE: Mean Squared Error
  • • RMSE: Root Mean Squared Error
  • • MAPE: Mean Absolute Percentage Error

Goodness of Fit

  • • R²: Coefficient of Determination
  • • Adjusted R²: Corrected for model complexity
  • • AIC/BIC: Information criteria
  • • Cross-validation scores

Best Practices for Model Evaluation

Data Splitting

  • • Use appropriate train/test splits
  • • Consider stratified sampling
  • • Implement cross-validation
  • • Avoid data leakage

Metric Selection

  • • Choose metrics matching your objective
  • • Consider class imbalance
  • • Use multiple complementary metrics
  • • Understand metric limitations

Interpretation

  • • Analyze confusion matrices visually
  • • Check residual plots for patterns
  • • Consider confidence intervals
  • • Validate on independent datasets