Converting a dataframe to dictionary with multiple values

Question

I have a dataframe like

Sr.No   ID       A         B          C         D
 1     Tom     Earth    English      BMW
 2     Tom     Mars     Spanish      BMW               


        

Answer 1

Grouping by 'ID' and apply to_dict to each group with orient='list' comes pretty close:

df.groupby('ID').apply(lambda dfg: dfg.to_dict(orient='list')).to_dict()
Out[25]: 
{'John': {'A': ['Venus', nan],
  'B': ['Portugese', 'German'],
  'C': ['Mercedes', 'Audi'],
  'D': ['Blue', 'Red'],
  'ID': ['John', 'John'],
  'Sr.No': [4, 5]},
 'Michael': {'A': ['Mercury'],
  'B': ['Hindi'],
  'C': ['Audi'],
  'D': ['Yellow'],
  'ID': ['Michael'],
  'Sr.No': [3]},
 'Tom': {'A': ['Earth', 'Mars'],
  'B': ['English', 'Spanish'],
  'C': ['BMW', 'BMW'],
  'D': [nan, 'Green'],
  'ID': ['Tom', 'Tom'],
  'Sr.No': [1, 2]}}

It should just be a matter of formatting the result slightly.

Edit: to remove 'ID' from the dictionaries:

df.groupby('ID').apply(lambda dfg: dfg.drop('ID', axis=1).to_dict(orient='list')).to_dict()
Out[5]: 
{'John': {'A': ['Venus', nan],
  'B': ['Portugese', 'German'],
  'C': ['Mercedes', 'Audi'],
  'D': ['Blue', 'Red'],
  'Sr.No': [4, 5]},
 'Michael': {'A': ['Mercury'],
  'B': ['Hindi'],
  'C': ['Audi'],
  'D': ['Yellow'],
  'Sr.No': [3]},
 'Tom': {'A': ['Earth', 'Mars'],
  'B': ['English', 'Spanish'],
  'C': ['BMW', 'BMW'],
  'D': [nan, 'Green'],
  'Sr.No': [1, 2]}}

Answer 2

You can use groupby with orient of to_dict as list and convert the resultant series to a dictionary.

df.set_index('Sr.No', inplace=True)
df.groupby('ID').apply(lambda x: x.to_dict('list')).reset_index(drop=True).to_dict()

{0: {'C': ['Mercedes', 'Audi'], 'ID': ['John', 'John'], 'A': ['Venus', nan],  
     'B': ['Portugese', 'German'], 'D': ['Blue', 'Red']}, 
 1: {'C': ['Audi'], 'ID': ['Michael'], 'A': ['Mercury'], 'B': ['Hindi'], 'D': ['Yellow']}, 
 2: {'C': ['BMW', 'BMW'], 'ID': ['Tom', 'Tom'], 'A': ['Earth', 'Mars'], 
     'B': ['English', 'Spanish'], 'D': [nan, 'Green']}}

---

Inorder to remove ID, you can also do:

df.groupby('ID')['A','B','C','D'].apply(lambda x: x.to_dict('list'))  \
                                 .reset_index(drop=True).to_dict()