使用 pandas 的堆叠条形图

Question

我想以条形图的形式表示我的数据，如我的预期输出所示。

time,date,category
0,2002-05-01,2
1,2002-05-02,0
2,2002-05-03,0
3,2002-05-04,0
4,2002-05-05,0
5,2002-05-06,0
6,2002-05-07,0
7,2002-05-08,2
8,2002-05-09,2
9,2002-05-10,0
10,2002-05-11,2
11,2002-05-12,0
12,2002-05-13,0
13,2002-05-14,2
14,2002-05-15,2
15,2002-05-16,2
16,2002-05-17,2
17,2002-05-18,2
18,2002-05-19,0
19,2002-05-20,0
20,2002-05-21,1
21,2002-05-22,2
22,2002-05-23,0
23,2002-05-24,1
24,2002-05-25,0
25,2002-05-26,0
26,2002-05-27,0
27,2002-05-28,0
28,2002-05-29,1
29,2002-05-30,0

import pandas as pd
from datetime import datetime
import matplotlib.pyplot as plt

df = pd.read_csv('df.csv')
daily_category = df[['date','category']]
daily_category['weekday'] = pd.to_datetime(daily_category['date']).dt.day_name()
daily_category_plot = daily_category[['weekday','category']]

daily_category_plot[['category']].groupby('weekday').count().plot(kind='bar', legend=None)
plt.show()

但是，我收到以下错误

回溯（最后一次调用）：文件“day_plot.py”，第 10 行，位于 daily_category_plot[['category']].groupby('weekday').count().plot(kind='bar', legend=None) 文件“/home/..../.local/lib/python3.6/site-packages/pandas/core/frame.py”，第 6525 行，在 groupby 中 dropna=dropna, 文件“/home/..../.local/lib/python3.6/site-packages/pandas/core/groupby/groupby.py”，第 533 行，在 init 中 dropna=self.dropna, 文件“/home/..../.local/lib/python3.6/site-packages/pandas/core/groupby/grouper.py”，第 786 行，在 get_grouper 提高 KeyError(gpr) 键错误：'weekday'

********** 下面是我手动提取数据的进一步示例 returns 几乎是预期的输出，只是日期表示为数字而不是工作日名称。 ***********

Day,category1,category2,category3
Sunday,0,0,4
Monday,0,0,4
Tuesday,1,1,2
Wednesday,1,4,0
Thursday,0,2,3
Friday,1,1,2
Saturday,0,2,2

import pandas as pd 
import numpy as np 
import matplotlib.pyplot as plt

df = pd.read_csv('df.csv')

ax = df.plot.bar(stacked=True, color=['green', 'red', 'blue'])
ax.set_xticklabels(labels=df.index, rotation=70, rotation_mode="anchor", ha="right")
ax.set_xlabel('')
ax.set_ylabel('Number of days')
plt.show()

测试输出

更新代码生成奇怪的情节

import pandas as pd
from datetime import datetime
import matplotlib.pyplot as plt

df = pd.read_csv('df.csv')
daily_category = df[['time','date','category']]
daily_category['weekday'] = pd.to_datetime(daily_category['date']).dt.day_name()

ans = (daily_category.groupby(['weekday', 'category']) 
         .size()
         .reset_index(name='sum')
         .pivot(index='weekday', columns='category', values='sum')
      )

ans.plot.bar(stacked=True)
plt.show()

更新输出

Answer 1

import pandas as pd
import matplotlib.pyplot as plt

d = """0,2002-05-01,2  1,2002-05-02,0  2,2002-05-03,0  3,2002-05-04,0  4,2002-05-05,0  5,2002-05-06,0  6,2002-05-07,0  7,2002-05-08,2  8,2002-05-09,2  9,2002-05-10,0  10,2002-05-11,2  11,2002-05-12,0  12,2002-05-13,0  13,2002-05-14,2  14,2002-05-15,2  15,2002-05-16,2  16,2002-05-17,2  17,2002-05-18,2  18,2002-05-19,0  19,2002-05-20,0  20,2002-05-21,1  21,2002-05-22,2  22,2002-05-23,0  23,2002-05-24,1  24,2002-05-25,0  25,2002-05-26,0  26,2002-05-27,0  27,2002-05-28,0  28,2002-05-29,1  29,2002-05-30,0"""
df = pd.DataFrame([v.split(',') for v in d.split('  ')], columns=['time', 'date', 'category'])
df.time, df.category = df.time.astype(int), df.category.astype(int)

data = df.copy()
data['weekday'] = pd.to_datetime(data['date']).dt.day_name()
data.drop(columns=['time', 'date'], inplace=True)

weekdays = ['Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday', 'Sunday']
categories = sorted(list(set(df.category)))
counts = pd.DataFrame(0, index=weekdays, columns=categories)
for weekday, category in zip(data.weekday, data.category):
    counts.loc[weekday, category] += 1

counts.plot.bar(stacked=True);

Answer 2

此解决方案对列使用 groupby，并使用 pivot 转换返回的 Dataframe。这可以由 plot.bar() 绘制，但标签错误。因此索引被更改。

我确实复制并通过了你的代码，并通过

获得了一个 DataFrame

import pandas as pd
from io import StringIO
t = """time,date,category
0,2002-05-01,2
..."""
df = pd.read_csv(StringIO(t))
df['weekday'] = df.date.apply(lambda x: pd.to_datetime(x).weekday())

为了检查周三柱的预期输出，我使用了过滤器选项。

>>>df[df['weekday']==2]
     time        date  category  weekday
0      0  2002-05-01         2        2
7      7  2002-05-08         2        2
14    14  2002-05-15         2        2
21    21  2002-05-22         2        2
28    28  2002-05-29         1        2

所以我想在星期三只看类别 1 (1/5) 和类别 2 (4/5)。

ans = (df.groupby(["weekday", "category"]) 
         .size()
         .reset_index(name="sum")
         .pivot(index='weekday', columns='category', values='sum')
      )
ans.index = ['Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday', 'Sunday']
ans.plot.bar(stacked=True)

使用 pandas 的堆叠条形图

Stacking bar plot using pandas

python

matplotlib

dataframe

pandas