如何用Python绘制双平台因子对比散点图（多子图/单文件）

Test Data Preparation #

First, generate synthetic test data matching your requirements (100 samples, 9 pairs of metrics):

import pandas as pd
import numpy as np

# Set random seed for reproducibility
np.random.seed(42)

# Define 9 metric pairs
metrics = ['yield', 'mean', 'std', 'max', 'min', 'median', 'q1', 'q3', 'count']

# Create DataFrame with NP and Arg platform metrics
data = {f'{metric}_NP': np.random.normal(loc=50, scale=10, size=100) for metric in metrics}
data.update({f'{metric}_Arg': np.random.normal(loc=52, scale=9, size=100) for metric in metrics})
df = pd.DataFrame(data)

1. Combined Subplots in Single Figure #

This code creates a 3×3 grid of scatter plots to compare all metric pairs in one figure:

import matplotlib.pyplot as plt

# Set up figure and subplot grid
fig, axes = plt.subplots(nrows=3, ncols=3, figsize=(15, 12))
axes = axes.flatten()  # Convert to 1D array for easy iteration

# Define distinct colors for each platform
platform_colors = {'NP': '#1f77b4', 'Arg': '#ff7f0e'}

# Iterate through each metric pair
for idx, metric in enumerate(metrics):
    ax = axes[idx]
    
    # Plot NP vs Arg values for each sample
    ax.scatter(df[f'{metric}_NP'], df[f'{metric}_Arg'], color=platform_colors['NP'], alpha=0.6, label='NP vs Arg')
    
    # Add diagonal reference line (perfect correlation)
    min_val = min(df[f'{metric}_NP'].min(), df[f'{metric}_Arg'].min())
    max_val = max(df[f'{metric}_NP'].max(), df[f'{metric}_Arg'].max())
    ax.plot([min_val, max_val], [min_val, max_val], 'k--', alpha=0.5)
    
    # Set axis ranges with padding
    ax.set_xlim(min_val - 5, max_val + 5)
    ax.set_ylim(min_val - 5, max_val + 5)
    
    # Configure labels and title
    ax.set_xlabel(f'{metric}_NP')
    ax.set_ylabel(f'{metric}_Arg')
    ax.set_title(f'{metric} Comparison')
    ax.legend()

# Adjust layout to avoid overlapping elements
plt.tight_layout()
plt.savefig('combined_metric_comparison.png', dpi=100)
plt.show()

2. Individual Scatter Plots (Saved Separately) #

This code generates and saves each metric comparison as an independent file:

for metric in metrics:
    plt.figure(figsize=(8, 6))
    
    # Plot data points
    plt.scatter(df[f'{metric}_NP'], df[f'{metric}_Arg'], color=platform_colors['NP'], alpha=0.6, label='NP vs Arg')
    
    # Add reference diagonal line
    min_val = min(df[f'{metric}_NP'].min(), df[f'{metric}_Arg'].min())
    max_val = max(df[f'{metric}_NP'].max(), df[f'{metric}_Arg'].max())
    plt.plot([min_val, max_val], [min_val, max_val], 'k--', alpha=0.5)
    
    # Set axis ranges
    plt.xlim(min_val - 5, max_val + 5)
    plt.ylim(min_val - 5, max_val + 5)
    
    # Configure plot labels and title
    plt.xlabel(f'{metric}_NP')
    plt.ylabel(f'{metric}_Arg')
    plt.title(f'{metric}: NP vs Arg Platform Comparison')
    plt.legend()
    
    # Save to file and close figure
    plt.savefig(f'{metric}_np_vs_arg.png', dpi=100)
    plt.close()

Key Details #

• Axis Range Control: Calculates minimum and maximum values for each metric, then adds padding to ensure all data points are visible.
• Color Differentiation: Uses high-contrast blue and orange to clearly distinguish platform comparisons.
• Reference Line: The dashed diagonal line helps quickly assess how closely each sample's values align between the two platforms.
• Reproducibility: The fixed random seed ensures test data remains consistent across runs.

内容的提问来源于stack exchange，提问作者user3214212