[ Python 常見問題 ] Seaborn - create a heatmap of two categorical variables

 Source From Here

Question
I have the following datasets of three variables:

view plaincopy to clipboardprint?

1. df['Score'] Float dummy (1 or 0)  
2. df['Province'] an object column where each row is a region  
3. df['Product type'] an object indicating the industry.  

I would like to create a jointplot where on the x axis I have the different industries, on the y axis the different provinces and as colours of my jointplot I have the relative frequency of the score. Something like this. https://seaborn.pydata.org/examples/hexbin_marginals.html

HowTo
If you are looking for a heatmap, you could use Seaborn heatmap function. However, you need to pivot your table first.

Just creating a small example:

view plaincopy to clipboardprint?

1. import numpy as np   
2. import pandas as pd  
3. import seaborn as sns  
4. import matplotlib.pyplot as plt  
5.   
6. score = [1, 1, 1, 0, 1, 0, 0, 0]  
7. provinces = ['Place1' ,'Place2' ,'Place2', 'Place3','Place1', 'Place2','Place3','Place1']  
8. products = ['Product1' ,'Product3' ,'Product2', 'Product2','Product1', 'Product2','Product1','Product1']  
9. df = pd.DataFrame({'Province': provinces,  
10.                    'Product type': products,  
11.                    'score': score  
12.                   })  

My df looks like:

view plaincopy to clipboardprint?

1.    'Province''Product type''score'  
2. 0   Place1    Product1      1  
3. 1   Place2    Product3      1  
4. 2   Place2    Product2      1  
5. 3   Place3    Product2      0  
6. 4   Place1    Product1      1  
7. 5   Place2    Product2      0  
8. 6   Place3    Product1      0  
9. 7   Place1    Product1      0  

Then:

view plaincopy to clipboardprint?

1. df_heatmap = df.pivot_table(values='score',index='Province',columns='Product type',aggfunc=np.mean)  
2. sns.heatmap(df_heatmap,annot=True)  
3. plt.show()  

The result is: