Confusion Matrix Heatmap — lets-plot

A specialized heatmap visualization for evaluating classification model performance, displaying the counts or proportions of predicted vs actual class labels. The confusion matrix reveals true positives, false positives, true negatives, and false negatives at a glance, making it essential for understanding model behavior, identifying class imbalances, and diagnosing specific misclassification patterns.

Confusion Matrix Heatmap rendered with lets-plot

Python source (lets-plot)

""" anyplot.ai
confusion-matrix: Confusion Matrix Heatmap
Library: letsplot 4.9.0 | Python 3.13.13
Quality: 86/100 | Updated: 2026-05-09
"""

import os

import numpy as np
import pandas as pd
from lets_plot import (
    LetsPlot,
    aes,
    coord_fixed,
    element_blank,
    element_rect,
    element_text,
    geom_text,
    geom_tile,
    ggplot,
    ggsave,
    ggsize,
    labs,
    scale_fill_gradient,
    theme,
    theme_minimal,
)


LetsPlot.setup_html()

# Theme tokens (see prompts/default-style-guide.md)
THEME = os.getenv("ANYPLOT_THEME", "light")
PAGE_BG = "#FAF8F1" if THEME == "light" else "#1A1A17"
ELEVATED_BG = "#FFFDF6" if THEME == "light" else "#242420"
INK = "#1A1A17" if THEME == "light" else "#F0EFE8"
INK_SOFT = "#4A4A44" if THEME == "light" else "#B8B7B0"

# Data - Multi-class classification results for image classifier
np.random.seed(42)

class_names = ["Cat", "Dog", "Bird", "Fish"]
n_classes = len(class_names)

# Create a realistic confusion matrix with strong diagonal
# and some realistic misclassification patterns
confusion_data = np.array(
    [
        [45, 8, 3, 2],  # Cat: sometimes confused with Dog
        [6, 52, 4, 1],  # Dog: sometimes confused with Cat
        [2, 3, 38, 5],  # Bird: sometimes confused with Fish
        [1, 2, 7, 41],  # Fish: sometimes confused with Bird
    ]
)

# Build long-form data for geom_tile
rows = []
for i, true_label in enumerate(class_names):
    for j, pred_label in enumerate(class_names):
        count = confusion_data[i, j]
        rows.append(
            {"True Label": true_label, "Predicted Label": pred_label, "Count": count, "true_idx": i, "pred_idx": j}
        )

df = pd.DataFrame(rows)

# Calculate percentages for annotation (row normalization = recall)
total_per_row = confusion_data.sum(axis=1, keepdims=True)
percentages = (confusion_data / total_per_row * 100).astype(int)
df["Percentage"] = [percentages[r["true_idx"], r["pred_idx"]] for _, r in df.iterrows()]
df["Label"] = df.apply(lambda r: f"{r['Count']}\n({r['Percentage']}%)", axis=1)

# Set category order for proper matrix layout
df["True Label"] = pd.Categorical(df["True Label"], categories=class_names[::-1], ordered=True)
df["Predicted Label"] = pd.Categorical(df["Predicted Label"], categories=class_names, ordered=True)

# Determine text color based on count (theme-adaptive)
max_count = df["Count"].max()
if THEME == "light":
    df["text_color"] = df["Count"].apply(lambda c: "#1A1A17" if c > max_count * 0.4 else "#4A4A44")
else:
    df["text_color"] = df["Count"].apply(lambda c: "#F0EFE8" if c > max_count * 0.4 else "#B8B7B0")

# Color gradient for sequential data (Blues)
if THEME == "light":
    low_color = "#E7F0F9"
    high_color = "#08519C"
else:
    low_color = "#1F3D5C"
    high_color = "#74B3E5"

# Create confusion matrix heatmap
plot = (
    ggplot(df, aes(x="Predicted Label", y="True Label", fill="Count"))
    + geom_tile(color=INK_SOFT, size=1.5, tooltips="none")
    + geom_text(aes(label="Label", color="text_color"), size=14, fontface="bold")
    + scale_fill_gradient(low=low_color, high=high_color, name="Count")
    + labs(x="Predicted Label", y="True Label", title="confusion-matrix · letsplot · anyplot.ai")
    + theme_minimal()
    + theme(
        plot_background=element_rect(fill=PAGE_BG, color=PAGE_BG),
        panel_background=element_rect(fill=PAGE_BG),
        plot_title=element_text(size=28, face="bold", color=INK),
        axis_title=element_text(size=22, color=INK),
        axis_text=element_text(size=18, color=INK_SOFT),
        legend_background=element_rect(fill=ELEVATED_BG, color=INK_SOFT),
        legend_title=element_text(size=18, color=INK),
        legend_text=element_text(size=14, color=INK_SOFT),
        panel_grid=element_blank(),
    )
    + ggsize(1200, 1200)
    + coord_fixed()
)

# Save as PNG (scale 3x for 3600x3600 px)
ggsave(plot, f"plot-{THEME}.png", path=".", scale=3)

# Save interactive HTML
ggsave(plot, f"plot-{THEME}.html", path=".")

Part of Confusion Matrix Heatmap on anyplot.ai.

Other implementations