How to make a correlation heatmap with seaborn but filtering by a specific column?

user_13408947 0
Viewed 1457

I want to make a correlation heatmap between 2 columns in a df, but filtered by a3 columns. I have a DF like this with 3 columns, and I want to make 1 heatmap to each region, but I couldn't find the way to make it.

chldmort    adfert      region
34.75       7.300000    Africa
122.75      111.699997  Americas
60.25       52.099998   Asia
170.50      124.800003  Europe
168.50      18.600000   Oceania

I try to this in seaborn, but I couldn't get an efficient way to do it for all-region in a row.

tmp=df.loc[:,['chldmort','adfert','region']].dropna()
tmp_africa=tmp[tmp['region']=='Africa']
tmp_americas=tmp[tmp['region']=='Americas']
tmp_asia=tmp[tmp['region']=='Asia']
tmp_europe=tmp[tmp['region']=='Europe']
tmp_oceania=tmp[tmp['region']=='Oceania']
plt.figure(figsize=(20,20))
plt.subplot(5,1,1)
plt.title("chldmort over adfert, grouped by Africa",size=15)
sns.heatmap(tmp_africa.corr(cmap='Reds', annot=True, vmax=.99, vmin=0.60, linewidths=0.9)
plt.subplot(5,1,2)
sns.heatmap(tmp_americas.corr(), cmap='Reds', annot=True, vmax=.99, vmin=0.60, linewidths=0.9)
plt.title("chldmort over adfert, grouped by Americas",size=15)
plt.subplot(5,1,3)
plt.title("chldmort over adfert, grouped by Asia",size=15)
sns.heatmap(tmp_asia.corr(), cmap='Reds', annot=True, vmax=.99, vmin=0.60, linewidths=0.9)
plt.subplot(5,1,4)
plt.title("chldmort over adfert, grouped by Europe",size=15)
sns.heatmap(tmp_europe.corr(), cmap='Reds', annot=True, vmax=.99, vmin=0.60, linewidths=0.9)
plt.subplot(5,1,5)
plt.title("chldmort over adfert, grouped by Oceania",size=15)
sns.heatmap(tmp_oceania.corr(), cmap='Reds', annot=True, vmax=.99, vmin=0.60, linewidths=0.9);
seaborn heatmap python
1 Answers

I made some editions to improve the plotting. You could try something like this:

fig = plt.figure(figsize = (30,50))
cont=1
for i in list(set(df.iloc[:,2])):         #2 because we are getting the values of the column 2(region)
      reg = tmp.loc[tmp['region'] == i]   #we get the dataframe filter by region
      reg= reg.iloc[:,[0,1]]              #we get the columns of chldmort and adfert 

      ax1 = fig.add_subplot(5, 1, cont)   #we are adding a subplot to the general figure
      ax1.title.set_text("chldmort over adfert, grouped by "+i)               #we set the title of the of ax1 with the current region name

      sns.heatmap(reg.corr(), ax=ax1, cmap='Reds', annot=True, vmax=.99, vmin=0.60, linewidths=0.9)  
      #By doing ax=ax1, we are assigning the subplot ax1 the current heatmap figure

      cont=cont+1

You can checkout more info about ploting multiple heatmaps here. And, for helpful info about corr(), you can click these links: heatmap-correlations, corr().

user_136762020
Related