Pandas：Pivot 类型化数据集

Question

Suppose I have data in the following format:

Time, Type, Subtype, Value
0,A,Ab,1
0,A,Ac,2
0,B,Ba,1

And I need:

Time, Type-A-Ab-Value, Type-A-Ac-Value, Type-B-Ba-Value
0, 1, 2, 1

Is there a pandas primitive that will solve this problem in one stroke?

Answer 1

Use DataFrame.set_index to set the multilevel index then use DataFrame.unstack on the levels 1, 2 to reshape the dataframe, then use .map to flatten the MultiIndex columns:

df1 = df.set_index(['Time', 'Type', 'Subtype']).unstack(level=[1, 2])
df1.columns = df1.columns.map(lambda s: 'Type-' + '-'.join(s[1:]) + '-Value')

OR, it is also possible to use DataFrame.pivot :

df['pvt'] = 'Type-' + df['Type'] + '-' + df['Subtype'] + '-Value'
df1 = df.pivot('Time', 'pvt', 'Value').rename_axis(columns=None)

Result:

print(df1)

      Type-A-Ab-Value  Type-A-Ac-Value  Type-B-Ba-Value
Time                                                   
0                   1                2                1