如何仅绘制 seaborn 热图的下三角形？

Question

我有一个列表列表all_genres，其中每个子列表包含特定电视节目的流派，每个不同的子列表指的是不同的电视节目。我想绘制一个 seaborn 热图，但只有较低的三角形部分。这是我迄今为止尝试过的：

import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
sns.set(rc={'figure.figsize':(16,9)})

all_genres = [['Drama', 'History'], ['Documentary'], ['Action', 'Drama', 'History'], ['Documentary'], ['Crime', 'Drama', 'Thriller'], ['Action', 'Adventure', 'Drama'], ['Crime', 'Drama', 'Thriller'], ['Documentary'], ['Documentary'], ['Documentary'], ['Animation', 'Adventure', 'Comedy'], ['Crime', 'Drama'], ['Documentary', 'History', 'War'], ['Animation', 'Action', 'Adventure'], ['Documentary'], ['Documentary', 'Music'], ['Crime', 'Drama', 'Mystery'], ['Documentary', 'History', 'War'], ['Fantasy', 'Horror', 'Mystery'], ['Documentary'], ['Drama'], ['Animation', 'Action', 'Adventure'], ['Adventure', 'Drama', 'Sci-Fi'], ['Documentary', 'History', 'War'], ['Biography', 'Drama', 'History'], ['Crime', 'Drama', 'Mystery'], ['Crime', 'Drama', 'Thriller'], ['Animation', 'Action', 'Adventure'], ['Animation', 'Crime', 'Drama'], ['Comedy', 'News', 'Talk-Show'], ['Documentary'], ['Animation', 'Action', 'Adventure'], ['Drama', 'Sci-Fi', 'Thriller'], ['Documentary'], ['Animation', 'Action', 'Adventure'], ['Comedy'], ['Comedy', 'Romance'], ['Drama', 'Romance'], ['Documentary'], ['Comedy'], ['Drama', 'Thriller', 'War'], ['Comedy'], ['Documentary', 'History', 'War'], ['Drama', 'Fantasy', 'Horror'], ['Animation', 'Adventure', 'Comedy'], ['Crime', 'Drama', 'Fantasy'], ['Comedy'], ['Animation', 'Adventure', 'Drama'], ['Comedy', 'Drama'], ['Comedy', 'Drama'], ['Biography', 'Crime', 'Drama'], ['Comedy', 'Drama'], ['Biography', 'Drama', 'History'], ['Comedy'], ['Drama'], ['Crime', 'Drama'], ['Comedy', 'War'], ['Drama', 'Mystery', 'Sci-Fi'], ['Action', 'Comedy', 'Drama'], ['Action', 'Drama', 'History'], ['Animation', 'Action', 'Adventure'], ['Comedy'], ['Crime', 'Drama', 'Thriller'], ['Animation', 'Comedy'], ['Crime', 'Drama', 'Mystery'], ['Crime', 'Thriller'], ['Animation', 'Action', 'Adventure'], ['Comedy'], ['Animation', 'Comedy'], ['Drama'], ['Drama', 'Mystery'], ['Comedy', 'Music'], ['Documentary', 'Comedy'], ['Comedy', 'Drama'], ['Animation', 'Action', 'Adventure'], ['Crime', 'Drama', 'History'], ['Animation', 'Action', 'Adventure'], ['Crime', 'Drama', 'Mystery'], ['Animation', 'Comedy', 'Drama'], ['Adventure', 'Drama', 'Western'], ['Documentary', 'Talk-Show'], ['Documentary', 'Crime'], ['Action', 'Adventure', 'Drama'], ['Drama', 'Horror', 'Mystery'], ['Crime', 'Drama'], ['Comedy'], ['Comedy', 'Drama'], ['Crime', 'Drama', 'Mystery'], ['Drama', 'History'], ['Crime', 'Drama', 'Mystery'], ['Crime', 'Drama', 'Mystery'], ['Crime', 'Drama', 'Thriller'], ['Comedy', 'Talk-Show'], ['Drama', 'Romance'], ['Animation', 'Crime', 'Drama'], ['Crime', 'Drama', 'Mystery'], ['Crime', 'Drama', 'Mystery'], ['Comedy'], ['Animation', 'Action', 'Adventure'], ['Animation', 'Action', 'Comedy'], ['Action', 'Drama', 'History'], ['Crime', 'Mystery', 'Thriller'], ['Comedy', 'Talk-Show'], ['Action', 'Adventure', 'Mystery'], ['Animation', 'Short', 'Comedy'], ['Action', 'Crime', 'Drama'], ['Comedy'], ['Comedy', 'Drama', 'Romance'], ['Animation', 'Action', 'Adventure'], ['Documentary', 'Crime'], ['Comedy', 'Drama', 'Music'], ['Comedy'], ['Drama', 'Romance'], ['Action', 'Crime', 'Drama'], ['Animation', 'Action', 'Adventure'], ['Drama', 'History', 'War'], ['Crime', 'Drama', 'Thriller'], ['Drama'], ['Action', 'Comedy'], ['Animation', 'Comedy'], ['Documentary', 'Adventure'], ['Documentary'], ['Comedy', 'Drama'], ['Animation', 'Action', 'Adventure'], ['Comedy', 'Game-Show'], ['Drama', 'Sport'], ['Adventure', 'Drama', 'Family'], ['Comedy'], ['Comedy'], ['Crime', 'Drama', 'Mystery'], ['Comedy'], ['Adventure', 'Comedy', 'Drama'], ['Family', 'Fantasy', 'Music'], ['Comedy', 'Drama'], ['Drama', 'Romance'], ['Animation', 'Action', 'Adventure'], ['Drama'], ['Action', 'Crime', 'Drama'], ['Comedy', 'Music', 'Musical'], ['Comedy'], ['Animation', 'Action', 'Drama'], ['Documentary'], ['Crime', 'Drama', 'Mystery'], ['Drama', 'War'], ['Animation', 'Action', 'Adventure'], ['Crime', 'Drama', 'Mystery'], ['Comedy', 'Drama'], ['Documentary', 'Biography', 'Crime'], ['Crime', 'Drama', 'History'], ['Comedy'], ['Animation', 'Action', 'Adventure'], ['Animation', 'Action', 'Adventure'], ['Biography', 'Drama', 'History'], ['Comedy', 'Talk-Show'], ['Comedy', 'Sci-Fi'], ['Comedy'], ['Crime', 'Drama'], ['Comedy', 'Crime', 'Drama'], ['Comedy', 'Drama'], ['Comedy', 'Romance'], ['Animation', 'Action', 'Adventure'], ['Crime', 'Drama', 'Mystery'], ['Action', 'Adventure', 'Crime'], ['Animation', 'Action', 'Adventure'], ['Comedy'], ['Animation', 'Action', 'Adventure'], ['Crime', 'Drama', 'Thriller'], ['Comedy', 'Family'], ['Documentary', 'History', 'War'], ['Biography', 'Drama'], ['Comedy'], ['Comedy', 'Musical', 'Sci-Fi'], ['Animation', 'Comedy', 'Drama'], ['Documentary', 'Sport'], ['Comedy', 'Game-Show'], ['Action', 'Adventure', 'Drama'], ['Comedy', 'Reality-TV'], ['Animation', 'Action', 'Adventure'], ['Drama', 'Romance'], ['Crime', 'Drama', 'Horror'], ['Comedy'], ['Comedy', 'Drama', 'Family'], ['Crime', 'Drama', 'Thriller'], ['Crime', 'Drama', 'Mystery'], ['Animation', 'Drama', 'Fantasy'], ['Action', 'Adventure', 'Biography'], ['Comedy', 'Drama'], ['Drama', 'Fantasy', 'Romance'], ['Action', 'Crime', 'Mystery'], ['Crime', 'Drama', 'Thriller'], ['Crime', 'Drama', 'Mystery'], ['Crime', 'Drama', 'Thriller'], ['Drama', 'Mystery', 'Sci-Fi'], ['Drama', 'Sci-Fi', 'Thriller'], ['Animation', 'Comedy', 'Sci-Fi'], ['Crime', 'Drama', 'Mystery'], ['Drama', 'Thriller'], ['Comedy'], ['Drama'], ['Comedy'], ['Animation', 'Action', 'Crime'], ['Biography', 'Crime', 'Drama'], ['Biography', 'Drama', 'History'], ['Comedy'], ['Action', 'Adventure', 'Biography'], ['Action', 'Comedy', 'Fantasy'], ['Crime', 'Drama'], ['Crime', 'Drama'], ['Crime', 'Drama'], ['Comedy', 'Drama', 'Romance'], ['Animation', 'Action', 'Adventure'], ['Animation', 'Action', 'Family'], ['Drama'], ['Animation', 'Comedy', 'Sci-Fi'], ['Action', 'Drama', 'History'], ['Crime', 'Drama', 'Mystery'], ['Crime', 'Drama'], ['Comedy', 'News', 'Talk-Show'], ['Animation', 'Action', 'Adventure'], ['Animation', 'Action', 'Adventure'], ['Drama', 'Mystery', 'Thriller'], ['Comedy', 'News', 'Talk-Show'], ['Comedy', 'Fantasy', 'Horror'], ['Drama', 'Fantasy', 'Mystery'], ['Adventure', 'Fantasy'], ['Animation', 'Action', 'Adventure'], ['Comedy', 'Crime'], ['Crime', 'Drama'], ['Comedy', 'Drama'], ['Drama', 'Mystery', 'Sci-Fi'], ['Comedy', 'Drama'], ['Comedy', 'Talk-Show'], ['Drama'], ['Animation', 'Family', 'Comedy'], ['Comedy', 'Horror'], ['Comedy', 'Drama'], ['Drama'], ['Documentary', 'Short'], ['Crime', 'Drama', 'Mystery'], ['Comedy', 'Romance'], ['Comedy', 'Family', 'Music'], ['Drama', 'Mystery', 'Sci-Fi'], ['Comedy', 'Crime']]
unique_genres = sorted({i for j in all_genres for i in j})
genre_matrix = []
for i in unique_genres:
    temp = []
    for j in unique_genres:
        s = sum(1  if i in m and j in m and i!=j else 0 for m in all_genres)
        temp.append(s)
    genre_matrix.append(temp)
genre_matrix = np.array(genre_matrix)

ax = sns.heatmap(genre_matrix, xticklabels=unique_genres, yticklabels=unique_genres, annot=True, linewidths=.42, cbar=True, cbar_kws={'label': 'Colorbar'})
#plt.savefig('seaborn_cross_fold_plot.png', bbox_inches="tight", pad_inches=0)
plt.show()

生成的图片是这样的：

我只想获取此热图的下部或底部，因为左下部分和右上部分相同，即它们只是对角线的反射。我想要得到的结果图片是这样的：

如何在 seaborn 或 matplotlib 中做到这一点？

Answer 1

派对有点晚了！... 另一种方式

corr= df_new.corr()

# Getting the Upper Triangle of the co-relation matrix
matrix = np.triu(corr)

# using the upper triangle matrix as mask 
sns.heatmap(corr, annot=True, mask=matrix)

Answer 2

您可以查看https://seaborn.pydata.org/examples/many_pairwise_correlations.html 并找出答案

mask = np.zeros_like(corr, dtype=np.bool)
mask[np.triu_indices_from(mask)] = True

将是用于屏蔽矩阵上三角部分的代码。然后您将 sns.heatmap 的代码更改为以下代码（如网站建议的那样）：

sns.heatmap(corr, mask=mask, cmap=cmap, vmax=.3, center=0,
            square=True, linewidths=.5, cbar_kws={"shrink": .5})

之后，您可以根据需要自定义其他属性。

Answer 3

你可以重复使用这个例子：

import seaborn as sns
import matplotlib.pyplot as plt

df = sns.load_dataset('car_crashes')
corr =  df.corr()

# Create a mask
mask = np.triu(np.ones_like(corr, dtype=bool))

# Create a custom divergin palette
cmap = sns.diverging_palette(100, 7, s=75, l=40,
                            n=5, center="light", as_cmap=True)

plt.figure(figsize=(10, 6))
sns.heatmap(corr, mask=mask, center=0, annot=True,
            fmt='.2f', square=True, cmap=cmap)

plt.show();

如果要显示主对角线，可以添加：

np.fill_diagonal(mask, False)

here 的积分