Python实现：为热图类别列设专属底色并按数值调深浅

我需要制作一款热图，要求：

• X轴的5个类别各有专属的整体底色
• 每个月份对应的数值决定该单元格颜色的深浅

目前使用sns.heatmap绘制，现有代码如下，通过cmap="Greys"能呈现数值差异，但希望实现每列不同预设底色且不添加额外单元格：

sns.heatmap(viewed, linewidth=.5, cmap="Greys", fmt="d")

解决方案 #

方法一：底色叠加灰度数值层 #

通过先绘制列专属底色，再叠加基于数值的灰度层，实现底色固定、数值控制深浅的效果：

import seaborn as sns
import matplotlib.pyplot as plt
import numpy as np

# 替换为你的实际数据矩阵，形状为(月份数, 类别数)
viewed = np.random.randint(10, 100, (6, 5))
# 定义5个类别对应的专属底色
base_colors = ["#FF9999", "#99FF99", "#9999FF", "#FFCC99", "#CC99FF"]

fig, ax = plt.subplots(figsize=(8, 4))

# 绘制列专属底色
for col in range(viewed.shape[1]):
    # 生成对应列的全1矩阵，用于填充底色
    col_base = np.ones((viewed.shape[0], 1))
    ax.imshow(col_base, 
              extent=[col, col+1, viewed.shape[0], 0],
              cmap=plt.cm.colors.ListedColormap([base_colors[col]]),
              alpha=1)

# 归一化数值到0-1区间，用于控制灰度层透明度
norm = plt.Normalize(viewed.min(), viewed.max())
normed_data = norm(viewed)
# 反转灰度逻辑：数值越大，灰度越浅，底色显示越清晰
gray_data = 1 - normed_data

# 叠加数值灰度层
ax.imshow(gray_data,
          cmap="Greys",
          extent=[0, viewed.shape[1], viewed.shape[0], 0],
          alpha=gray_data,
          interpolation="nearest")

# 设置刻度与边框
ax.set_xticks(np.arange(viewed.shape[1]) + 0.5)
ax.set_xticklabels(["类别1", "类别2", "类别3", "类别4", "类别5"])
ax.set_yticks(np.arange(viewed.shape[0]) + 0.5)
ax.set_yticklabels(["1月", "2月", "3月", "4月", "5月", "6月"])
ax.grid(color="white", linewidth=0.5)

plt.tight_layout()
plt.show()

方法二：列专属渐变色彩映射 #

为每个类别创建独立的渐变色彩映射，数值直接控制该底色的深浅程度：

import seaborn as sns
import matplotlib.pyplot as plt
from matplotlib.colors import LinearSegmentedColormap
import numpy as np

# 替换为你的实际数据
viewed = np.random.randint(10, 100, (6, 5))
# 定义5个类别的基础底色
base_colors = ["#FF9999", "#99FF99", "#9999FF", "#FFCC99", "#CC99FF"]

fig, ax = plt.subplots(figsize=(8, 4))

# 为每个类别生成专属渐变色彩映射（浅到深）
custom_cmaps = []
for color in base_colors:
    # 生成基础色的浅色调（低数值）和深色调（高数值）
    light_shade = plt.cm.colors.to_rgba(color, alpha=0.3)
    dark_shade = plt.cm.colors.to_rgba(color, alpha=1.0)
    cmap = LinearSegmentedColormap.from_list(f"col_cmap_{color}", [light_shade, dark_shade])
    custom_cmaps.append(cmap)

# 逐列绘制热图
for col_idx in range(viewed.shape[1]):
    col_data = viewed[:, col_idx].reshape(-1, 1)
    # 绘制当前列，使用专属色卡
    sns.heatmap(col_data,
                ax=ax,
                cmap=custom_cmaps[col_idx],
                fmt="d",
                cbar=False,
                linewidths=0.5,
                xticklabels=[f"类别{col_idx+1}"],
                # 仅第一列显示Y轴刻度，避免重复
                yticklabels=["1月", "2月", "3月", "4月", "5月", "6月"] if col_idx == 0 else False,
                extent=[col_idx, col_idx+1, viewed.shape[0], 0])

plt.tight_layout()
plt.show()

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