GnuplotEx.ML.Confusion (gnuplot_ex v0.2.4)

Confusion matrix heatmap visualization.

Provides functions to visualize classification confusion matrices as heatmaps with proper labels, normalization options, and customizable color schemes.

Examples

# Basic confusion matrix
matrix = [
  [50, 2, 1],
  [3, 45, 2],
  [1, 1, 48]
]
classes = ["Cat", "Dog", "Bird"]

plot = GnuplotEx.ML.Confusion.plot(matrix, classes)
GnuplotEx.render(plot, :svg)

# Normalized confusion matrix
plot = GnuplotEx.ML.Confusion.plot(matrix, classes,
  normalize: true,
  title: "Normalized Confusion Matrix"
)

Summary

Functions

accuracy(matrix)

Calculate accuracy from confusion matrix.

f1_score(matrix)

Calculate F1 score from confusion matrix.

normalize_matrix(matrix)

Normalize confusion matrix by row (true class).

plot(matrix, class_names, opts \\ [])

Plot a confusion matrix as a heatmap.

precision(matrix)

Calculate per-class precision from confusion matrix.

recall(matrix)

Calculate per-class recall from confusion matrix.

Functions

accuracy(matrix)

Calculate accuracy from confusion matrix.

Returns the overall accuracy (sum of diagonal / sum of all elements).

Example

iex> matrix = [[50, 10], [5, 35]]
iex> GnuplotEx.ML.Confusion.accuracy(matrix)
0.85

f1_score(matrix)

Calculate F1 score from confusion matrix.

Returns a list of F1 scores for each class.

F1 = 2 (precision recall) / (precision + recall)

Example

iex> matrix = [[50, 10], [5, 35]]
iex> GnuplotEx.ML.Confusion.f1_score(matrix)
[0.8771929824561403, 0.823529411764706]

normalize_matrix(matrix)

Normalize confusion matrix by row (true class).

Each row sums to 1.0, showing the proportion of predictions for each true class.

Example

iex> matrix = [[50, 10], [5, 35]]
iex> GnuplotEx.ML.Confusion.normalize_matrix(matrix)
[[0.8333333333333334, 0.16666666666666666], [0.125, 0.875]]

plot(matrix, class_names, opts \\ [])

Plot a confusion matrix as a heatmap.

Creates a heatmap visualization of a confusion matrix with class labels on both axes and optional normalization.

Parameters

  • matrix - 2D list representing the confusion matrix (rows = true labels, cols = predicted labels)
  • class_names - List of class names for labels

Options

  • :normalize - Normalize by row (true class) when true (default: false)
  • :title - Plot title (default: "Confusion Matrix")
  • :palette - Color palette (default: :viridis)
  • :show_values - Annotate cells with values (default: false, gnuplot limitation)
  • :colorbar - Show colorbar (default: true)
  • :value_format - Format string for normalized values (default: "%.2f")

Examples

# Binary classification
matrix = [[95, 5], [10, 90]]
classes = ["Negative", "Positive"]
plot = GnuplotEx.ML.Confusion.plot(matrix, classes)

# Multi-class with normalization
matrix = [[50, 2, 1], [3, 45, 2], [1, 1, 48]]
classes = ["Cat", "Dog", "Bird"]
plot = GnuplotEx.ML.Confusion.plot(matrix, classes,
  normalize: true,
  palette: :plasma
)

# Custom color scheme
plot = GnuplotEx.ML.Confusion.plot(matrix, classes,
  palette: {:custom, ["#ffffff", "#ff0000"]}
)

precision(matrix)

Calculate per-class precision from confusion matrix.

Precision = True Positives / (True Positives + False Positives)

Returns a list of precision values for each class.

Example

iex> matrix = [[50, 10], [5, 35]]
iex> GnuplotEx.ML.Confusion.precision(matrix)
[0.9090909090909091, 0.7777777777777778]

recall(matrix)

Calculate per-class recall from confusion matrix.

Recall = True Positives / (True Positives + False Negatives)

Returns a list of recall values for each class.

Example

iex> matrix = [[50, 10], [5, 35]]
iex> GnuplotEx.ML.Confusion.recall(matrix)
[0.8333333333333334, 0.875]