Matplotlib：在单个图上绘制两列或更多列的计数图

Question

My dataFrame, df:

Sno | Attribute_1   | Attribute_2   | Attribute_3
__________________________________________________
1   | option_1      | option_3      |option_2
2   | option_1      | option_1      |option_1
3   | option_2      | option_2      |option_2
4   | option_1      | option_1      |option_3
5   | option_3      | option_2      |option_2
6   | option_3      | option_3      |option_1
7   | option_1      | option_3      |option_2

Here Attribute_1, Attribute_2 and Attribute_3 contains categorical data - option_1 or option_2 or option_3 for each of the rows.

I want to create a count plot on the same plot for all the attributes. I am able to do it for one column by:

sns.countplot(x="Attribute_1", data=df);

I can individually create for each of the attributes, but what I am looking for it that on the same plot I can have count plot for all the attributes. ie X-axis will have attributes, and each attribute will have three count plot.

Answer 1

Seaborn usually works best with long form datasets. Ie instead of 3 columns with different options for each attribute you would have two columns, one for the options and one for the attributes. This can easily be created via pd.melt . Then the hue value can be used on the "options" column:

sns.countplot(x="variable", hue="value", data=pd.melt(df))

Complete example:

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

a= np.random.choice(["option_{}".format(i) for i in [1,2,3]], size=(12,3))
df = pd.DataFrame(a, columns=["Attribute_{}".format(i) for i in list("ABC")])

sns.countplot(x="variable", hue="value", data=pd.melt(df))

plt.show()

Equally you can interchange x and hue :

sns.countplot(x="value", hue="variable", data=pd.melt(df))