如何根据年-周为数据框中的每一列执行词袋模型？

Question

 Year-Week  Job_list   Comments_2
    2019-01    Doc-1      Doc-1
    2019-01    Doc-2      Doc-2
    2019-01    Doc-3      Doc-3
    2019-02    Doc-4      Doc-4
    2019-03    Doc-5      Doc-5

I want the output separately for each column according to year-week after applying bag of words model

Answer 1

You can use .groupby to group the output according to year-week.

After that you can use .apply(list) to transform the grouped output into a list.

df = pd.DataFrame([['2019-01', 'Doc-1', 'Doc-1'], ['2019-01', 'Doc-2', 'Doc-2'],['2019-01','Doc-3','Doc-3'],['2019-02','Doc-4','Doc-4'],['2019-03','Doc-5','Doc-5']],columns= ['Year-Week', 'Job_list', 'Comments_2'])

#Use groupby and then .apply(list)
job_list_grouped = df.groupby('Year-Week')['Job_list'].apply(list)
print(job_list_grouped)

The output looks like that:

Year-Week
2019-01    [Doc-1, Doc-2, Doc-3]
2019-02                  [Doc-4]
2019-03                  [Doc-5]

You could simply do the same for the other column. And from there you can use it to transform it to anything you need.

EDIT:

You can use the module Counter from the library collections for that!

Here is my new code:

import pandas as pd
from collections import Counter

df = pd.DataFrame([['2019-01', 'Doc-1', 'Doc-1'], ['2019-01', 'Doc-2', 'Doc-2'],['2019-01','Doc-3','Doc-3'],['2019-02','Doc-4','Doc-4'],['2019-03','Doc-5','Doc-5']],columns= ['Year-Week', 'Job_list', 'Comments_2'])

job_list_grouped = df.groupby('Year-Week')['Job_list'].apply(list).apply(Counter)
print(job_list_grouped)
print(job_list_grouped.to_dict())

Note how I only added another apply to the end of the groupby function.

The first print puts out:

Year-Week
2019-01    {'Doc-1': 1, 'Doc-2': 1, 'Doc-3': 1}
2019-02                            {'Doc-4': 1}
2019-03                            {'Doc-5': 1}

If you need to have this in a dictionary format you can simply add to_dict() to do this:

{'2019-01': Counter({'Doc-1': 1, 'Doc-2': 1, 'Doc-3': 1}), '2019-02': Counter({'Doc-4': 1}), '2019-03': Counter({'Doc-5': 1})}

Don´t worry about the Counter around your dictionaries. It still behaves exactly like a dict .