如何在 seaborn 传说中 select 和排序值

Question

我只想在 seaborn 热图图例中包含某些值。具体来说，我有一个 "nan" 类别，我不想在图例中看到它。

我正在尝试将医院患者的病房移动绘制为一种分类热图，不同的颜色代表不同的病房。我借用了这段代码 heatmap-like plot, but for categorical variables in seaborn 来为热图配置我的输入 table。空白单元格表示患者在这些日期不在医院。

import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
import matplotlib.patches as mpatches
import pandas as pd

data = {'12/3': [np.nan, 'Ward_B', np.nan],
        '13/3': [np.nan, 'Ward_B', np.nan],
        '14/3': [np.nan, 'Ward_B', 'ED'],
        '15/3': ['ED', 'Ward_A', 'Ward_C'],
        '16/3': ['ED', 'Ward_A', 'Ward_C'],
        '17/3': ['Ward_A', 'Ward_A', 'Ward_C'],
        '18/3': ['Ward_A', np.nan, 'Ward_C'],
        '19/3': ['Ward_A', np.nan, 'Ward_A'],
        '20/3': [np.nan, np.nan, 'Ward_A']}

df = pd.DataFrame (data, columns = ['12/3',
                                    '13/3',
                                    '14/3',
                                    '15/3',
                                    '16/3',
                                    '17/3',
                                    '18/3',
                                    '19/3',
                                    '20/3'])

# Create dataframe of patient IDs
patient_codes_df = pd.DataFrame(['Patient_A', 'Patient_B', 'Patient_C'])
# change heading
patient_codes_df = patient_codes_df.rename(columns={0:'Patient'})
# Merge
df2 = pd.concat([patient_codes_df, df], axis=1)
# Make Patient column the index
df3 = df2.set_index('Patient')
df3

df3 是我的输入数据的样子。

这就是我绘制热图的方式

value_to_int = {j:i for i,j in enumerate(pd.unique(df3.values.ravel()))}
n = len(value_to_int)

cmap = sns.color_palette("Accent", n) # set colours

fig, ax = plt.subplots(1, 1, figsize = (6, 2), dpi=300)

mask = df3.isnull()
ax = sns.heatmap(df3.replace(value_to_int), cmap=cmap, mask=mask, linewidths=0.1, linecolor='#b5b5b5') 

ax.set_ylabel('')

# modify colorbar:
colorbar = ax.collections[0].colorbar 
r = colorbar.vmax - colorbar.vmin 
colorbar.set_ticks([colorbar.vmin + r / n * (0.5 + i) for i in range(n)])
colorbar.set_ticklabels(list(value_to_int.keys()))  
plt.xticks(rotation=90)
plt.show()

我想从图例中删除 "nan"，并重新排序，使其按照合理的顺序排列，如 ED、Ward_A、Ward_B、Ward_C.

感谢您的帮助。

Answer 1

定义 value_to_int 时必须删除 nans。你借用的代码很好，但我想更直接的方法是在字典中手动定义你的颜色，然后在绘图时用这个替换你的data.frame：

lvls = {'ED': 0, 'Ward_A': 1, 'Ward_B': 2, 'Ward_C': 3}
cmap = sns.color_palette("Accent", len(lvls)) # set colours

fig, ax = plt.subplots(1, 1, figsize = (6, 2), dpi=300)
sns.heatmap(df3.replace(lvls),cmap=cmap,mask=df3.isnull(),linewidths=.1,
            linecolor='#b5b5b5',ax=ax)

colorbar = ax.collections[0].colorbar 
r = colorbar.vmax - colorbar.vmin
n = len(lvls)
colorbar.set_ticks([colorbar.vmin + r / n * (0.5 + i) for i in range(n)])
colorbar.set_ticklabels(list(lvls.keys()))
plt.xticks(rotation=90)
plt.show()

Answer 2

所以一种有效的方法是删除颜色条并手动创建我自己的图例：

# Define colours
cols = ["#ffff99", '#beaed4', '#fdc692', '#7fc97f', '#fd4396']

# Transform categorical variables into numbers for heatmap
value_to_int = {j:i for i,j in enumerate(pd.unique(df3.values.ravel()))}
n = len(value_to_int)
cmap = sns.color_palette(cols, n) # set colours

# Plot figure
fig, ax = plt.subplots(1, 1, figsize = (8, 3), dpi=300)
sns.set(font_scale=1.27, style='whitegrid')
mask = df3.isnull()
ax = sns.heatmap(df3.replace(value_to_int),
                 cmap=cmap, mask=mask, linewidths=0.1, linecolor='#b5b5b5',
                 cbar=False, # Remove the color bar legend
                 xticklabels=2)

# Tweaking figire
ax.set_ylabel('')
plt.xticks(rotation=90)

# Create a new legend
ED_patch = mpatches.Patch(color='#beaed4', label='ED')
A_patch = mpatches.Patch(color='#7fc97f', label='Ward A')
B_patch = mpatches.Patch(color='#fd4396', label='Ward B')
C_patch = mpatches.Patch(color='#3da1bf', label='Ward C')

ax.legend(handles=[ED_patch,A_patch,B_patch,C_patch],
         bbox_to_anchor=(1.22, 1),
         prop={'size': 12})

如何在 seaborn 传说中 select 和排序值

How to select and order values in seaborn legend

python

matplotlib

heatmap

seaborn