Threshold Metrics Explorer

This component computes classification performance metrics across varying threshold levels for a binary classifier's predicted probabilities.

Given a table containing predicted probabilities and true class labels, the component iteratively evaluates the following metrics at each threshold from 0 to 1 (configurable step size):

• Precision

• Recall

• F1 Score

• Accuracy

It outputs a table mapping each threshold to its corresponding metric values. This enables model evaluation under varying decision thresholds, which is especially useful for applications where class imbalance or cost-sensitive classification exists.

Input ports

1. Type: Table
   Model results
   Table with original class column and predicted class column.

Output ports

1. Type: Table
   Output
   A table with rows for each threshold value and columns: threshold precision recall f1_score accuracy