pandas convert a nested dictionary to mutiIndex rows and columns

user_12975329 0
Viewed 179

I have a nested dictionary and i want to make it to a multiIndex rows and colums like below. But my data is lost in table somehow. 

    test= {12: {'Category 1': {'TestA': {'att_1': 1, 'att_2': 'whatever'}, 'TestB': {'att_1': 3, 'att_2': 'spring'}}, 'Category 2': {'TestA': {'att_1': 23, 'att_2': 'another'}, 'TestB': {'att_1': 9, 'att_2': 'summer'}}}, 15: {'Category 1': {'TestA': {'att_1': 10, 'att_2': 'foo'}, 'TestB': {'att_1': 29, 'att_2': 'fall'}}, 'Category 2': {'TestA': {'att_1': 30, 'att_2': 'bar'}, 'TestB': {'att_1': 36, 'att_2': 'winter'}}}}
columns=pd.MultiIndex.from_arrays([['TestA','TestA','TestB','TestB'],['att_1','att_2','att_1','att_2']])

The format I want:

              TestA       TestB      
              att_1 att_2 att_1 att_2
12 Category 1   NaN   NaN   NaN   NaN
   Category 2   NaN   NaN   NaN   NaN
15 Category 1   NaN   NaN   NaN   NaN
   Category 2   NaN   NaN   NaN   NaN

And I did 

    pd.DataFrame(test,index=pd.MultiIndex.from_arrays([[12,12,15,15],['Category 1','Category 2','Category 1','Category 2']]),columns=pd.MultiIndex.from_arrays([['TestA','TestA','TestB','TestB'],['att_1','att_2','att_1','att_2']]))

My data is lost as below:

             TestA       TestB      
              att_1 att_2 att_1 att_2
12 Category 1   NaN   NaN   NaN   NaN
   Category 2   NaN   NaN   NaN   NaN
15 Category 1   NaN   NaN   NaN   NaN
   Category 2   NaN   NaN   NaN   NaN

If I have only multiIndex rows, it would work, but I want both multiIndex rows and colums. 

     pd.DataFrame.from_dict({(i,j): test[i][j] 
                           for i in test.keys() 
                           for j in test[i].keys()},
                       orient='index')

                                           TestA                             TestB
12 Category 1  {'att_1': 1, 'att_2': 'whatever'}   {'att_1': 3, 'att_2': 'spring'}
   Category 2  {'att_1': 23, 'att_2': 'another'}   {'att_1': 9, 'att_2': 'summer'}
15 Category 1      {'att_1': 10, 'att_2': 'foo'}    {'att_1': 29, 'att_2': 'fall'}
   Category 2      {'att_1': 30, 'att_2': 'bar'}  {'att_1': 36, 'att_2': 'winter
multi-index pandas dataframe dictionary python
1 Answers

You can get the required data frame as:

import pandas as pd
import numpy as np

test= {12: {'Category 1': {'TestA': {'att_1': 1, 'att_2': 'whatever'}, 'TestB': {'att_1': 3, 'att_2': 'spring'}}, 'Category 2': {'TestA': {'att_1': 23, 'att_2': 'another'}, 'TestB': {'att_1': 9, 'att_2': 'summer'}}}, 15: {'Category 1': {'TestA': {'att_1': 10, 'att_2': 'foo'}, 'TestB': {'att_1': 29, 'att_2': 'fall'}}, 'Category 2': {'TestA': {'att_1': 30, 'att_2': 'bar'}, 'TestB': {'att_1': 36, 'att_2': 'winter'}}}}

# Row indexes
row_index = [[12,12,15,15],['Category 1','Category 2','Category 1','Category 2']]

# Column indexes
col_index = [['TestA','TestA','TestB','TestB'],['att_1','att_2','att_1','att_2']]

# Values row wise
values =[1,'whatever',3,'spring',23,'another',9,'summer',10,'foo',29,'fall',30,'bar',36,'winter']

# Convert the list of values to numpy array
value = np.array(values)

# Reshape the value as (4,4) array as the matrix/dataframe is of shape (4,4)
value = value.reshape(4,4)

# Get your required data frame
pd.DataFrame(value, index=row_index, columns=col_index)
user_113243580
Related