Reverse action of one hot vector in different features in dataframe-CodePudding

I had a dataframe with 2 categorical features like this:

Then use get_dummies to one hot vector these columns:

Now I want to get back from one hot vector to first columns, actually a reverse action of get_dummies. Is there any way to do this?

CodePudding user response：

Use DataFrame.melt for unpivot with filter 1 in DataFrame.query, then splitting varible column and reshape by DataFrame.set_index with Series.unstack:

df = pd.get_dummies(df1.astype(str))

df = df.melt(ignore_index=False).query('value == 1')
df[['a','b']] = df['variable'].str.rsplit('_', n=1, expand=True)
df = df.set_index('a', append=True)['b'].unstack().rename_axis(None, axis=1)

Or use DataFrame.stack with filter in Series.loc, convert multiIndex to DataFrame by MultiIndex.to_frame, splitting and pivoting by DataFrame.pivot:

df = df.stack().loc[lambda x: x.eq(1)].index.to_frame()
df[['a','b']] = df[1].str.rsplit('_', n=1, expand=True)
df = df.pivot(0,'a','b').rename_axis(index=None, columns=None)

CodePudding user response：

Use from_dummies (pandas 1.5 ):

df_original = pd.from_dummies(df_dummies, sep='_')

Output:

  EstateTypes AdverTypes
0           1          1
1           1          2
2           1          2
3           1          3
4           1          2