Python Pandas pivot_table 在透视后缺少列

Question

I have following data frame. The data frame is constructed by reading a csv file. Its a large data set but for this question purpose I have used 15 rows from the data set as an example.

   user_id   contrib_count   total_min_length     group_space     expert_level
0     23720        108           1112696               0             l-2
1     23720         13            442059               1             l-2
2     23720         12             32180               2             l-2
3     23720          2             20177               3             l-2
4     23720          1              1608              10             l-2
5   1265184         71            260186               0             l-G
6   1265184         10              3466               2             l-G
7   1265184          1             12081               4             l-G
8    513380        112           1049311               0             l-4
9    513380          1                97               1             l-4
10   513380        113            361980               2             l-4
11   513380         19           1198323               3             l-4
12   513380          2             88301               4             l-4
13    20251        705          17372707               0             l-G
14    20251        103           2327178               1             l-G

Expected Results After pivot what I want is following data frame:

group_space        0      1     2     3     4     5    6   7    8   9    10  expert_level
user_id
20251             705    103    68    24    18     2    6 NaN  NaN   5   22     l-G                                                                  
23720             108     13    12     2   NaN   NaN  NaN NaN  NaN NaN    1     l-2

Reason I am doing this is once I do this I can use this for a prediction task where expert_level as label data.

So far I have done following to to build the above matrix but I am unable to get the expert_level column as shown after the pivot.

This is what I have done:

class GroupAnalysis():

    def __init__(self):
        self.df = None
        self.filelocation = '~/somelocation/x.csv'

    def pivot_dataframe(self):

        raw_df = pd.read_csv(self.filelocation)
        self.df = raw_df[(raw_df['group_space'] < 11)]
        self.df.set_index(['user_id', 'group_space'], inplace=True)
        self.df = self.df['contrib_count'].unstack()

By doing this I get:

group_space        0      1     2     3     4     5    6   7    8   9    10
user_id
20251             705    103    68    24    18     2    6 NaN  NaN   5   22                                                                
23720             108     13    12     2   NaN   NaN  NaN NaN  NaN NaN    1 

As you can see I am missing the expert_level column at the end. So the question is How can I get above data frame with the expert_level as I shown in my "Expected Results"?

Answer 1

When you unstacked, you were only unstacking a series contrib_count - expert_level and total_min_length were already gone at that point.

Instead of setting index and unstacking, you can just use .pivot()

pivoted = df.pivot('user_id', 'group_space', 'contrib_count')

Then, create a frame with user_id as the index and expert_level as a column, getting rid of duplicates:

lookup = df.drop_duplicates('user_id')[['user_id', 'expert_level']]
lookup.set_index(['user_id'], inplace=True)

Then join your pivot and lookup

result = pivoted.join(lookup)

EDIT: If you also want to include total_min_length , you can do a second pivot:

pivoted2 = df.pivot('user_id', 'group_space', 'total_min_length')

and join all three instead of two:

result = pivoted.join(lookup).join(pivoted2, lsuffix="_contrib_count", rsuffix="_total_min_length")

Note that lsuffix and rsuffix are required to disambiguate columns, as both pivots have 0, 1, 2, 3, 4 , and 10 columns from your example data.