如何使用matplotlib在python中按组和构面网格绘制具有不同颜色的直方图

Question

我有以下数据：

import pandas as pd

data = pd.DataFrame({"group":   ["aa", "aa", "aa", "aa", "bb", "bb", "bb", "bb"],
                     "segment": ["da", "et", "da", "et", "da", "et", "da", "et"],
                     "country": ["br", "br", "th", "th", "br", "br", "th", "th"],
                     "N":       [31, 23, 17, 9, 4, 100, 10, 20],
                     "totalN":  [84, 84, 389, 389, 84, 84, 389, 389]}
                    )

我想在python中进行相同的绘图，例如以下R代码产生的绘图

ggplot(data, aes(x=segment, y=N, fill=group)) +
geom_bar(stat="identity") +
ggtitle("group") +
facet_grid(country~.)+
geom_text(aes(label=percent(round(N / totalN, 2))), position=position_stack(vjust=0.5), size=3) +
coord_flip()

我努力了

data_groupped = data.groupby(['group', 'segment'])
data_groupped.plot(x='segment', y='N', kind='hist')

它分别生成每个直方图。

因此，预期的输出是这样的：

Answer 1

使用熊猫图，您可以

选项1]使用pivot_table重塑groups数据

import matplotlib.pyplot as plt

groups = data.groupby('country')
fig, axes = plt.subplots(groups.ngroups,sharex=True)
for (g, grp), ax in zip(groups, axes.flatten()):
    grp_df = grp.pivot_table(index='segment', columns='group', values='N', aggfunc=np.sum)
    grp_df.plot.barh(stacked=True, ax=ax, sharex=True)

---

选项2]首先将数据重塑为df ，然后使用plot

df = (data.groupby('country')
        .apply(lambda x: x.groupby(['segment', 'group'])['N'].sum().unstack())
        .unstack(level=0)
        .reorder_levels((1,0), axis=1)
        .sort_index(axis=1)
)
cgroups = df.groupby(level=0, axis=1)
fig, axes = plt.subplots(cgroups.ngroups, sharex=True)
for (c, grp), ax in zip(cgroups, axes.flatten()):
    sp = grp[c].plot.barh(stacked=True, ax=ax, sharex=True)

df

---

选项3]如果不需要分离子图

df = (data.groupby('country')
        .apply(lambda x: x.groupby(['segment', 'group'])['N'].sum().unstack()))
df.plot.barh(stacked=True)

df