使用pandas可视化工具绘制多个子图

Question

我使用的是来自kaggle的IGN review数据集，我正在尝试获得每个任天堂平台的频率图x给定发布日期的星期几，代码如下

import pandas as pd
df = pd.read_csv("ign.csv")
datetime_df = pd.DataFrame({'year': df["release_year"],
                   'month': df["release_month"],
                   'day': df["release_day"]})
df["date"] = pd.to_datetime(datetime_df)

df["week_day"] = df["date"].apply(lambda x : x.weekday_name)

nintendo = ['Wii','Nintendo DS','Nintendo 3DS','Nintendo DS',
            'Game Boy', 'Game Boy Color','Nintendo 64DD','Game Boy Advance',
            'New Nintendo 3DS','GameCube','Nintendo DSi','Super NES']

base_nintendo = df[df["platform"].isin(nintendo)]

data = base_nintendo.groupby(["platform","week_day"]).size()

data =data.unstack().fillna(0).stack()

data

输出如下：

platform          week_day 
Game Boy          Friday         5.0
                  Monday         5.0
                  Saturday       0.0
                  Sunday         0.0
                  Thursday       0.0
                  Tuesday        4.0
                  Wednesday      8.0
Game Boy Advance  Friday       131.0
                  Monday       109.0
                  Saturday       0.0
                  Sunday         1.0
                  Thursday     153.0
                  Tuesday      123.0
                  Wednesday    106.0
Game Boy Color    Friday        89.0
                  Monday        43.0
                  Saturday       1.0
                  Sunday         1.0
                  Thursday      55.0
                  Tuesday       78.0
                  Wednesday     89.0
GameCube          Friday        99.0
                  Monday       100.0
                  Saturday       3.0
                  Sunday         0.0
                  Thursday      83.0
                  Tuesday      124.0
                  Wednesday    100.0

我试过这样做：

data.groupby("platform").plot("barh")

但它只给了我最后一个平台(wii)：

​

​

Answer 1

一种解决方案是使用seaborn和plot barh。

data = data.unstack().fillna(0).stack()
data = data.reset_index().rename(columns={0:'value'})

import seaborn as sns
import matplotlib.pyplot as plt
fig, ax = plt.subplots(figsize=(10,7))
sns.barplot(y='platform',x='value', hue='week_day', data=data, orient='h')
plt.show()

Answer 2

注意上面的图，每个组都有一条线，比如Super NES ....？这些是其他绘图的matplotlib.AxesSubplot对象。

groupby.plot实际上为您的每个组返回一个matplotlib.AxesSubplot对象。另一方面，ipython notebook只显示您的最后一幅图。

解决方案是:将data.groupby("platform").plot("barh")更改为my_axes = data.groupby("platform").plot("barh")，然后逐个处理它们，如下所示

for ax in my_axes:
    ax.savefig(filename)

或者，您可以执行以下操作：

gp = data.groupby("platform")
f, axes = plt.subplots(5, 5)  # or any other large enough subplot grid
for k, ax in zip(gp.groups, axes.ravel()):
    gp.get_group(k).plot('barh', ax=ax)