如何在柱状图中绘制带有多个数值的子类别

你可以用Python的matplotlib库来完成这个需求，下面是具体的实现示例，以两种常见的可视化方式为例：

方式1：箱线图（展示数值分布） #

适合直观呈现每个子类别下数值的分布特征：

import matplotlib.pyplot as plt
import numpy as np

# 你的数据字典
data = {'A': {'pos': [3, 5, 7], 'neg': [8, 10, 11]},
        'B': {'pos': [4, 7, 8], 'neg': [7, 10, 12]},
        'C': {'pos': [1, 3, 4], 'neg': [2, 8, 8]}}

# 提取数据
categories = list(data.keys())
pos_values = [data[cat]['pos'] for cat in categories]
neg_values = [data[cat]['neg'] for cat in categories]

# 绘图设置
fig, ax = plt.subplots(figsize=(8, 6))
box_width = 0.35
x = np.arange(len(categories))

# 绘制正负类别箱线图
ax.boxplot(pos_values, positions=x - box_width/2, widths=box_width, patch_artist=True, boxprops=dict(facecolor='#66b3ff'))
ax.boxplot(neg_values, positions=x + box_width/2, widths=box_width, patch_artist=True, boxprops=dict(facecolor='#ff9999'))

# 添加标注
ax.set_title('Category-wise Positive & Negative Values')
ax.set_ylabel('Value')
plt.xticks(x, categories)
ax.legend([plt.Rectangle((0,0),1,1,fc='#66b3ff'), plt.Rectangle((0,0),1,1,fc='#ff9999')], ['Positive', 'Negative'])
plt.tight_layout()
plt.show()

方式2：散点图（展示单个数值点） #

适合展示每个子类别下的所有具体数值，通过抖动避免点重叠：

import matplotlib.pyplot as plt
import numpy as np

data = {'A': {'pos': [3, 5, 7], 'neg': [8, 10, 11]},
        'B': {'pos': [4, 7, 8], 'neg': [7, 10, 12]},
        'C': {'pos': [1, 3, 4], 'neg': [2, 8, 8]}}

categories = list(data.keys())
x = np.arange(len(categories))
box_width = 0.35

# 整理散点数据
x_pos = np.repeat(x - box_width/2, [len(v) for v in pos_values])
y_pos = [val for sublist in pos_values for val in sublist]
x_neg = np.repeat(x + box_width/2, [len(v) for v in neg_values])
y_neg = [val for sublist in neg_values for val in sublist]

# 绘图
fig, ax = plt.subplots(figsize=(8, 6))
ax.scatter(x_pos + np.random.normal(0, 0.02, len(x_pos)), y_pos, color='#66b3ff', label='Positive')
ax.scatter(x_neg + np.random.normal(0, 0.02, len(x_neg)), y_neg, color='#ff9999', label='Negative')

# 添加标注
ax.set_title('Category-wise Positive & Negative Values')
ax.set_ylabel('Value')
plt.xticks(x, categories)
ax.legend()
plt.tight_layout()
plt.show()

你可以根据想要的最终可视化效果选择对应的实现方式。

内容的提问来源于stack exchange，提问作者Jozef