Seaborn.heatmap 不根据特定值调整颜色

Question

我想要一个混淆矩阵的热图，它根据每个 class 的百分比生成颜色。例如，最高百分比变为黑色，其他百分比根据它们的百分比变亮（更高的百分比 = 更深的颜色）。我尝试更改 vmin 和 vmax，但颜色会根据 'counts' 值而不是 grouped_percentages.

进行更改

categories = ['a', 'b', 'c']
group_percentages = []
counts = []
for i in range (len(cf)):
  for j in range(len(cf)):
    group_percentages.append(cf[j,i]/np.sum(cf[:,i]))
    counts.append(cf[j,i])

group_percentages = ['{0:.2%}'.format(value) for value in
                group_percentages]

counts = ['{0:0.0f}'.format(value) for value in
            counts]
labels = [f'{v1}\n{v2}' for v1, v2 in zip(group_percentages, counts)]
labels = np.asarray(labels).reshape(3,3,order='F')

sns.heatmap(cf, annot=labels, fmt='', xticklabels=categories, yticklabels=categories, cmap='Greys', vmax=100, cbar=False)

输出：

如您所见，虽然我将 vmax 设置为 100，但 cf[0,0] 为 100%，但热图中的颜色为灰色，但 cf[1,1] 为 89%，并且其颜色为黑色。

Answer 1

您应该使用 group_percentages 作为热图的数据，但首先您需要在 3x3 矩阵中重塑此列表：

percentages_matrix = np.reshape(group_percentages, (3, 3))

完整代码

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

cf = np.array([[23, 0, 3],
               [0, 106, 5],
               [0, 12, 76]])

categories = ['a', 'b', 'c']
group_percentages = []
counts = []
for i in range(len(cf)):
    for j in range(len(cf)):
        group_percentages.append(cf[j, i]/np.sum(cf[:, i]))
        counts.append(cf[j, i])

percentages_matrix = np.reshape(group_percentages, (3, 3))
group_percentages = ['{0:.2%}'.format(value) for value in group_percentages]

labels = [f'{v1}\n{v2}' for v1, v2 in zip(group_percentages, counts)]
labels = np.asarray(labels).reshape(3, 3, order = 'F')

sns.heatmap(percentages_matrix, annot = labels, fmt = '', xticklabels = categories, yticklabels = categories, cmap = 'Greys', vmax = 1, vmin = 0, cbar = False)

plt.show()

Seaborn.heatmap 不根据特定值调整颜色

Seaborn.heatmap doesn't adjust colors based on specific value

python

data-visualization

matplotlib

heatmap

seaborn

完整代码