使用Seaborn绘制热图时重命名熊猫数据框列

Question

I have a dataframe named recast with the following structure:

data = [[ 374000,np.nan,749500,np.nan],
            [np.nan,np.nan,298000,np.nan],
            [540065.326633,np.nan,904750,np.nan],
            [np.nan,np.nan,514000,np.nan],
            [411000,np.nan,np.nan,np.nan]]
cols = pd.MultiIndex.from_tuples([('EUR, Oil (bbls)',c) for c in ['DE WITT','FAYETTE','GONZALES','LAVACA']], names=('','County'))
index = ['1776 ENERGY','ALMS','BURLINGTON','BXP','CHESAPEAKE']

recast=pd.DataFrame(data,index=index,columns=cols)
recast.index.name = 'ShortName'

print(recast)

            EUR, Oil (bbls)                         
 County             DE WITT FAYETTE  GONZALES LAVACA
ShortName                                           
1776 ENERGY   374000.000000     NaN  749500.0    NaN
ALMS                    NaN     NaN  298000.0    NaN
BURLINGTON    540065.326633     NaN  904750.0    NaN
BXP                     NaN     NaN  514000.0    NaN
CHESAPEAKE    411000.000000     NaN       NaN    NaN

What I would like to do is make a heatmap that has only the County names and not the column names that it is giving me. When I call .info() on the dataframe I get:

<class 'pandas.core.frame.DataFrame'>
Index: 29 entries, 1776 ENERGY to VERDUN OIL
Data columns (total 4 columns):
(EUR, Oil (bbls), DE WITT)     14 non-null float64
(EUR, Oil (bbls), FAYETTE)     3 non-null float64
(EUR, Oil (bbls), GONZALES)    23 non-null float64
(EUR, Oil (bbls), LAVACA)      5 non-null float64
dtypes: float64(4)

So by default my x labels on the heatmap are, for example: EUR, Oil (bbls), DE WITT , when I only want: DE WITT .

I have tried elminating the unwanted part by trying rename the columns in the dataframe and by trying to set x_label, but can't get it to work. For plotting I am using the following code:

fig = plt.figure(71)
fig = sns.heatmap(recast, cmap='coolwarm', linewidths=0.25, linecolor='black')
fig = plt.xticks(rotation=0)
plt.tight_layout
plt.show()

Also, I would like to remove the X and Y titles(?) that appear beneath/left of the labels.

Answer 1

You can use .xs() to achieve your goal ( see documentation ).

With the following sample:

data = [[ 374000,np.nan,749500,np.nan],
        [np.nan,np.nan,298000,np.nan],
        [540065.326633,np.nan,904750,np.nan],
        [np.nan,np.nan,514000,np.nan],
        [411000,np.nan,np.nan,np.nan]]
cols = pd.MultiIndex.from_tuples([('EUR, Oil (bbls)',c) for c in ['DE WITT','FAYETTE','GONZALES','LAVACA']], names=('','County'))
index = ['1776 ENERGY','ALMS','BURLINGTON','BXP','CHESAPEAKE']

recast=pd.DataFrame(data,index=index,columns=cols)
recast.index.name = 'ShortName'

print(recast)
            EUR, Oil (bbls)                         
County              DE WITT FAYETTE  GONZALES LAVACA
ShortName                                           
1776 ENERGY   374000.000000     NaN  749500.0    NaN
ALMS                    NaN     NaN  298000.0    NaN
BURLINGTON    540065.326633     NaN  904750.0    NaN
BXP                     NaN     NaN  514000.0    NaN
CHESAPEAKE    411000.000000     NaN       NaN    NaN

print(recast.xs('EUR, Oil (bbls)',axis=1, drop_level=True))
#axis = 1 for column multi-index selection
#drop_level = True to drop 'EUR Oil (bbls)' column level

County             DE WITT  FAYETTE  GONZALES  LAVACA
ShortName                                            
1776 ENERGY  374000.000000      NaN  749500.0     NaN
ALMS                   NaN      NaN  298000.0     NaN
BURLINGTON   540065.326633      NaN  904750.0     NaN
BXP                    NaN      NaN  514000.0     NaN
CHESAPEAKE   411000.000000      NaN       NaN     NaN

To plot what you want I rewrite your plotting section with a minimal set:

_recast = recast.xs('EUR, Oil (bbls)',axis=1, drop_level=True)
fig = sns.heatmap(_recast, cmap='coolwarm', linewidths=0.25, linecolor='black')
fig.set_xlabel(' ')
fig.set_ylabel(' ')
plt.show()

Note that to remove X and Y labels you can simply rename them with fig.set_[x or y]label() functions: