panda dataframe 可视化数值列的不同 bin 中类别列的不同值的百分比

Question

I have a pandas dataframe with two columns col1 and class . class is binary. I want to plot a histogram and visualize the percentage of each one of class values on different bins of col1 column. Here are my attempts:

1- Two histograms, one for each value of class column:

df.col1[df.class == 0].hist()

df.col1[df.class == 1].hist()

2- Put them all (two values of class) together in one chart

df.groupby('class').col1.hist(alpha=0.9)

As you can see from the first two graphs, those rows with class==1 are rare comparing to another class==0 and when we put them together (third graph), we don't see their effect (look at those tiny orange areas in the chart). One solution is using the percentage of each value of class within each bin. I tried this one:

df.groupby('class').col1.transform(lambda x: x/sum(x)).hist(alpha=0.9)

and apparently didn't work. I'm looking for a way to visualize the percentage of each class value within different bins.

Answer 1

Since the number of items per class is highly unbalanced, there is no way to have both plots in the same y-axes for histograms, If both clases have similar distribution among the values you can use distplots to perform some normalization on the data:

uniques = df['class'].unique()
targets = [df.col1[df['class'] == val] for val in uniques]

for target in targets:
    sns.distplot(target, rug=True)