绘制分组的pandas数据帧

Question

I spent a few hours searching for an answer, but I can't seem to get one.

Long story short, I have a dataframe. The following code will produce the dataframe in question (albeit anonymised with random numbers):

variable1 = ["Attribute 1","Attribute 1","Attribute 1","Attribute 1","Attribute 1","Attribute 1","Attribute 2","Attribute 2",
         "Attribute 2","Attribute 2","Attribute 2","Attribute 2","Attribute 3","Attribute 3","Attribute 3","Attribute 3",
         "Attribute 3","Attribute 3","Attribute 4","Attribute 4","Attribute 4","Attribute 4","Attribute 4","Attribute 4",
         "Attribute 5","Attribute 5","Attribute 5","Attribute 5","Attribute 5","Attribute 5"]


variable2 = ["Property1","Property2","Property3","Property4","Property5","Property6","Property1","Property2","Property3",
         "Property4","Property5","Property6","Property1","Property2","Property3",
         "Property4","Property5","Property6","Property1","Property2","Property3","Property4",
         "Property5","Property6","Property1","Property2","Property3","Property4","Property5","Property6"]

number = [93,224,192,253,186,266,296,100,135,169,373,108,211,194,164,375,211,71,120,334,59,164,348,50,249,18,251,343,172,41]

bar = pd.DataFrame({"variable1":variable1, "variable2":variable2, "number":number})

bar_grouped = bar.groupby(["variable1","variable2"]).sum()

The outcome should look like:

And the second one:

I have been trying to plot them with a bar chart and having the Properties as the groups and the different Attributes as the bars. Similar to this (plotted in Excel manually though). I would prefer to do it in the grouped datafarme, as to be able to plot with different groupings without the need to reset the index each time.

I hope this is clear.

Any help on this is hugely appreciated.

Thanks! :)

Answer 1

I wouldn't bother creating your groupby result (since you aren't aggregating anything). This is a pivot

---

bar.pivot('variable2', 'variable1', 'number').plot(kind='bar')

plt.tight_layout()
plt.show()

---

---

If aggregation is required, you can still start with your bar and use pivot_table

bar.pivot_table(index='variable2', columns='variable1', values='number', aggfunc='sum')

Answer 2

Use unstack first:

bar_grouped['number'].unstack(0).plot(kind='bar')

[out]

Answer 3

Below code will do what you trying to establish :

import numpy as np
import matplotlib.pyplot as plt

# set width of bar
barWidth = 0.25
f = plt.figure(figsize=(15,8))

bars={}
bar_pos={}
for i,proprty in enumerate(bar_grouped.unstack().columns.droplevel(0).tolist()):
    bars[i] = bar_grouped.unstack()['number',proprty].tolist()
    if(i==0):
        bar_pos[i]=2*np.arange(len(bars1))
    else:
        bar_pos[i]=[x + barWidth for x in bar_pos[i-1]] 
    plt.bar(bar_pos[i], bars[i], width=barWidth, edgecolor='white', label=proprty, figure=f)

# Add xticks on the middle of the group bars
plt.xlabel('group', fontweight='bold')
plt.xticks([2*r + 2*barWidth for r in range(len(bars[0]))], bar_grouped.unstack().index.tolist())
# plt.figure(figsize=(10,5))

# Create legend & Show graphic
plt.legend(loc=0)
plt.show()

I took the solution from here and modified it to fit your need. Hope this helps!