How to represent two dimensional categorical data in a Bar Chart using Python

Question

I have a dataset which contains columns: 'Month', 'Category', and 'Profitability'. I used following to find the sum of 'Profitability' for each month and category.

q1=df.groupby(['Month','Category'])['Profitability'].sum()

Here is the result I got.

Month  Category   
1      Cosmetics       2685.9000
       First Aid       2128.0200
       Magazine         703.8900
       Supplements    37005.6200
       Toiletries      1893.0600
2      Cosmetics       2569.0600
       First Aid       3282.7850
       Magazine         679.1100
       Supplements    36647.8800
       Toiletries      1357.7500
3      Cosmetics       1350.7925
       First Aid       2238.3100
       Magazine         371.1200
       Supplements    21444.0900
       Toiletries      1226.1600

I want to represent them in a bar chart. What would be the best approach to visualize these categorical data?

What do you want your visualization to depict? - What do you want the viewer to see? What are you trying to highlight to the viewer? How about a stacked bar chart?- one bar for each category with the values from each month stacked ? Or do you want to show changes over time for each category? — wwii
– wwii, Commented May 23, 2020 at 14:17
Or maybe a bar for each month with the category values stacked? — wwii
– wwii, Commented May 23, 2020 at 14:24
@wwii I want to set Month in X axis and Profit in Y axis. Need to draw 5 bars for each month for each Month — Nipuna Dilhara
– Nipuna Dilhara, Commented May 23, 2020 at 14:30
A simple line graph is often preferable to a bar plot, especially if you want to show changes over time. You could use different colors to show the categories. — Arne
– Arne, Commented May 23, 2020 at 14:54

Valdi_Bo · Accepted Answer · 2020-05-23 16:31:35Z

A preparatory step is to convert your DataFrame with a single column and a MultiIndex into a DataFrame with "normal" index and a separate column for each category:

df2 = df.Profitability.unstack()

Or if source of your data is a Series (not a DataFrame), then run:

df2 = q1.unstack()

The result, better fit as a figure source, is:

Category  Cosmetics  First Aid  Magazine  Supplements  Toiletries
Month                                                            
1         2685.9000   2128.020    703.89     37005.62     1893.06
2         2569.0600   3282.785    679.11     36647.88     1357.75
3         1350.7925   2238.310    371.12     21444.09     1226.16

To draw the figure, the basic approach is to take linear scale of y axis. The code to draw it is:

ax = df2.plot.bar(rot=0)
ax.get_figure().suptitle(t='Profitability', fontsize=20)
ax.legend(bbox_to_anchor=(1.35, 1.0));

The last instruction "moves" the legend a bit to the right (compared to its default location), otherwise it would obscure some bars (try to draw without this instruction).

The result is:

Note however that bars for Suppelements are very high, compared to other categories, which are very low.

This is why I came up with the second solution, based on logatithmic scale of y axis:

ax = df2.plot.bar(rot=0, logy=True)
ax.get_figure().suptitle(t='Profitability', fontsize=20)
ax.legend(bbox_to_anchor=(1.1, 1.0))
yTicks = [1000, 3000, 10000, 30000]
yTickLabels = [ f'{i:,}' for i in yTicks ]
ax.set_yticks(yTicks)
ax.set_yticklabels(yTickLabels);

Last 4 instructions change the default ticks on y axis, as in my opinion they are better readable that the default ticks (try to draw a figure without these 4 lines for comparison).

The result is:

Now heigths of bars are easier to compare and y ticks are chosen roughly on tick(n-1) * 3 basis.

Edit

If you want to have the legend with a title, add title='Category' parameter to ax.legend(...) instruction, whichever of the above solutions you choose.

Collectives™ on Stack Overflow

How to represent two dimensional categorical data in a Bar Chart using Python

1 Answer 1

Edit

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Edit

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related