Multicolumn plot Seaborn and Matplotlib

Question

I have this table of data:

Date, City, Sales, PSale1, PSale2

Where PSale1, and PSale2 are predicted prices for sales.

The data set looks like this:

Date,City,Sales,PSale1,PSale2
01/01,NYC,200,300,178
01/01,SF,300,140,100
02/01,NYC,400,410,33
02/01,SF,42,24,60

I want to plot this data in seaborn, I was working on to plot a graph which will have date on the x-axis, and PSale1 and PSale2 as bars for each city on y-axis, but I am confused how to work with such a data.

As much as I know seaborn doesn't allow plotting two variables on any axis, how should I approach this situation?

Answer 1

• Convert the dataframe to a long format
• Plot using seaborn.FacetGrid

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

data = {'Date': ['01/01', '01/01', '02/01', '02/01'],
        'City': ['NYC', 'SF', 'NYC', 'SF'],
        'Sales': [200, 300, 400, 42],
        'PSale1': [300, 140, 410, 24],
        'PSale2': [178, 100, 33, 60]}

df = pd.DataFrame(data)

# convert to long
dfl = df.set_index(['Date', 'City', 'Sales']).stack().reset_index().rename(columns={'level_3': 'pred', 0: 'value'})

# plot
g = sns.FacetGrid(dfl, row='City')
g.map(sns.barplot, 'Date', 'value', 'pred').add_legend()
plt.show()

---

• All in one figure

# shape the dataframe
dfc = df.drop(columns=['Sales']).set_index(['Date', 'City']).stack().unstack(level=1)
dfc.columns.name = None

# plot
dfc.plot.bar(stacked=True)
plt.xlabel('Date: Predicted')
plt.show()