我有一个包含399行(Words)和5列(Dates)的数据集。我想用热图可视化一些信息。我使用以下方法创建了一个枢轴表:
pd.pivot_table(df, index='Words', columns='Date', values='frequency', aggfunc=np.sum)输出:
Date 2018-02-18 2018-02-19 2018-02-20 2018-02-21 2018-02-22
Words
A NaN NaN NaN 2.0 2.0
B NaN NaN NaN NaN 1.0
C NaN NaN NaN NaN 1.0
D NaN 1.0 NaN NaN NaN
E NaN NaN 1.0 NaN NaN
... ... ... ... ... ...
RRR NaN 10.0 NaN NaN 90.0
SSS NaN 3.0 3.0 3.0 NaN
TTT NaN 4.0 NaN NaN NaN
UUU NaN NaN NaN 1.0 NaN
VVV NaN NaN NaN NaN 1.0
ZZZ NaN NaN 1.0 NaN 1.0
399 rows × 5 columns然后,我尝试使用以下代码行创建热图:
piv = pd.pivot_table(df, values="frequency",index=["Words"], columns=["Date"], fill_value=0)
ax = sns.heatmap(piv, square=False)但是,输出只显示这399行中的20行。是否有可能将热图中的所有行可视化?如果这是不可能的,我如何才能只看到最受欢迎的行(即在时间/日期中频率更高的行)?
你的帮助将不胜感激。谢谢。
发布于 2020-04-26 21:29:57
您的输出确实显示了所有的行,但是y标签减少了,因为它们会重叠太多,并且无法读取。
如果没有频率列,则可以创建一个列,并将所有值设置为1和df['frequency'] = 1。然后,聚合函数将对所有内容进行汇总。
您可以对piv数据进行排序,并使用idx = piv.sum(axis=1).sort_values(ascending=False).head(10).index获取10个最高值。然后,piv.loc[idx]将按该顺序得到这些行。
下面的代码显示了这些步骤。它还旋转滴答标签,以使他们更好地可读性在这种情况下。
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
N = 1000
abc = list('ABCDEFGHIJKLMNOPQRS')
df = pd.DataFrame({'Date':[f'2018-02-{i:02d}' for i in np.random.randint(18, 23, N)],
'Words': [abc[i]+abc[j] for i,j in np.random.randint(0, len(abc), (N, 2)) ] ,
'frequency': np.random.randint(1, 10, N)
})
# df['frequency'] = 1 # in case there wasn't a frequency column yet
piv = pd.pivot_table(df, values="frequency",index=["Words"], columns=["Date"], fill_value=0, aggfunc=np.sum)
idx = piv.sum(axis=1).sort_values(ascending=False).head(10).index
ax = sns.heatmap(piv.loc[idx], square=False)
ax.set_xticklabels(ax.get_xticklabels(), rotation=0) # rotate the x labels to be horizontally again
ax.set_yticklabels(ax.get_yticklabels(), rotation=0) # rotate the y labels to be horizontally
plt.show()

PS:显示所有的滴答声,以及所有的标签(它们可能过于拥挤)按字母顺序排列:
from matplotlib.ticker import FixedLocator
idx = piv.sort_values('Words', ascending=True).index
ax = sns.heatmap(piv.loc[idx], square=False)
ax.yaxis.set_major_locator(FixedLocator(np.arange(0.5, len(idx) + 0.5, 1)))
ax.set_yticklabels(idx, rotation=0, fontsize=6)或者看到标签左右交替(为了适应双轴),第二轴可能会有帮助:
ax.yaxis.set_major_locator(FixedLocator(np.arange(0.5, len(idx) + 0.5, 2)))
ax.set_yticklabels(idx[::2], rotation=0, fontsize=6)
secax = ax.secondary_yaxis('right')
secax.yaxis.set_major_locator(FixedLocator(np.arange(1.5, len(idx) + 0.5, 2)))
secax.set_yticklabels(idx[1::2], rotation=0, fontsize=6)https://stackoverflow.com/questions/61447388
复制相似问题