为Pyplot混淆矩阵热力图添加非传统侧边数值标注

Absolutely! This is totally achievable with Python's matplotlib and seaborn libraries—even if you're new to plotting, it’s manageable with a bit of step-by-step guidance. Here’s how you can add those custom error values to your confusion matrix heatmap:

Step 1: Set up your base heatmap #

First, let’s start with a basic confusion matrix heatmap (replace the example matrix with your actual data):

import matplotlib.pyplot as plt
import seaborn as sns

# Replace this with your real confusion matrix data
confusion_matrix = [
    [90, 0, 13],
    [6, 85, 0],
    [9, 0, 88]
]

# Create the core heatmap
plt.figure(figsize=(8, 6))
ax = sns.heatmap(confusion_matrix, annot=True, fmt='d', cmap='Blues', cbar=True)
plt.title("Confusion Matrix Heatmap")

Step 2: Add omission errors (13, 0, 0) above the top-edge squares #

We’ll use ax.text() to place these values right above each top cell, between the title and the heatmap:

# Coordinates for the center of each top cell (x-axis)
top_x_coords = [0.5, 1.5, 2.5]
# Position above the top edge of the heatmap
top_y_coord = ax.get_ylim()[1] + 0.1

# Your omission error values
omission_errors = [13, 0, 0]

# Loop to place each value
for x, val in zip(top_x_coords, omission_errors):
    ax.text(x, top_y_coord, str(val), ha='center', va='bottom', fontweight='bold')

Step 3: Add commission errors (0, 6, 9) to the right of the right-edge squares #

Similarly, we’ll place these values between the heatmap and the color legend:

# Coordinates for the center of each right-edge cell (y-axis; note matplotlib reverses y for heatmaps)
right_y_coords = [2.5, 1.5, 0.5]
# Position to the right of the heatmap's edge
right_x_coord = ax.get_xlim()[1] + 0.1

# Your commission error values
commission_errors = [0, 6, 9]

# Loop to place each value
for y, val in zip(right_y_coords, commission_errors):
    ax.text(right_x_coord, y, str(val), ha='left', va='center', fontweight='bold')

Step 4: Adjust layout and display the plot #

Finally, tweak the plot limits to make space for the new text and clean up the layout:

# Expand the plot area to fit the extra text
ax.set_xlim(ax.get_xlim()[0], ax.get_xlim()[1] + 0.5)
ax.set_ylim(ax.get_ylim()[0], ax.get_ylim()[1] + 0.5)

plt.tight_layout()
plt.show()

Quick notes for customization: #

• You can adjust the offset values (like +0.1 or +0.5) to move the text closer or farther from the heatmap.
• If your confusion matrix has more/fewer classes, just update the top_x_coords and right_y_coords lists to match the number of classes (e.g., for a 4x4 matrix, use [0.5, 1.5, 2.5, 3.5]).
• Change fontweight='bold' to other styles (like fontsize=12) if you want to tweak the text appearance.

内容的提问来源于stack exchange，提问作者SonicProtein