如何在 Pandas 中分组、计数然后绘制条形图？

Question

I have a Pandas dataframe that looks like the following.

year  month  class
----  -----  -----
2015  1      1
2015  1      1
2015  1      2
2015  1      2
...

I want to be able to create 2 bar chart series of of this data on one plot. If I can do a groupby , count and end up with a data frame then I am thinking I can just do a simple dataframe.plot.barh .

What I have tried is the following code.

x = df.groupby(['year', 'month', 'class'])['class'].count()

What x ends up being is a Series . So then I do the following to get a DataFrame .

df = pd.DataFrame(x)

Which gets me pretty close. The data ends up looking like the following.

clazz
year month clazz        
2015 1     1            2
     2     1           15
     2     2           45

But when I do a bar plot df.plot.bar() , I only see one series. The output desired is simply in one series, from 2015-01 to 2019-12, how many times did class 1 occur per month? And then another series, from 2015-01 to 2019-12, how many times did class 2 occur per month?

Any ideas on how to manipulate the data to be in this way?

Answer 1

A groupby - unstack should do the trick:

Data

df = pd.DataFrame([[2015, 1, 1],
                    [2015, 1, 1],
                    [2015, 1, 2],
                    [2015, 1, 2],
                    [2015, 1, 2],
                    [2015, 2, 1],
                    [2015, 2, 1],
                    [2015, 2, 1],
                    [2015, 2, 2],
                    [2015, 2, 2]], columns = ['year', 'month', 'class'])

Solution

df_gb = df.groupby(['year', 'month', 'class']).size().unstack(level=2)

Output

df_gb.plot(kind = 'bar')

Answer 2

We can also use DataFrame.pivot_table :

df.pivot_table(index=['year','month'],columns='class',aggfunc='size').plot(kind='bar')

---

or

df.pivot_table(index='class',columns=['year','month'],aggfunc='size').plot(kind='bar')