NormalizationType

Enums · AiDotNet.Evaluation.Enums

Specifies the normalization method for confusion matrices and metrics.

For Beginners

A raw confusion matrix shows counts (e.g., “50 true positives”). Normalized versions show proportions, which are easier to interpret:

  • None: Raw counts
  • ByTrue: “What percentage of actual positives were correctly identified?”
  • ByPredicted: “What percentage of predicted positives were correct?”
  • All: “What percentage of all samples fall into each cell?”

How It Works

Confusion matrix normalization helps interpret results, especially with imbalanced classes. Different normalizations answer different questions.

Fields

Field Summary
NormalizationType All Normalize by total: All cells sum to 1.
NormalizationType ByPredicted Normalize by predicted labels (columns): Each column sums to 1.
NormalizationType ByTrue Normalize by true labels (rows): Each row sums to 1.
NormalizationType Log Log normalization: Apply log transformation.
NormalizationType MinMax Min-max normalization: Scale values to [0, 1] range.
NormalizationType None No normalization: Raw counts in confusion matrix.
NormalizationType Robust Robust normalization: Use median and IQR instead of mean and std.
NormalizationType ZScore Z-score (standard) normalization: Transform to mean=0, std=1.
int value__