如何从 Python dataframe 中提取和分离随机元组值？

Question

Two values, say subject and subject category exist as columns in a data frame. Along with this I have the weight of subject in another column. I wish to create another data-frame that has random instances of a subject and it's corresponding subject category based on the weights of the subject. The tricky part here for me is to use both values ( subject, and subject category) together with the weights. While I am able to extract the value of the tuple and generate a random instance of the tuple based on the weight. I am not able to separate the tuple into its constituent elements to insert into the final data frame. The 'Zip' function is not working.

In my output dataframe I would want a Serial Number, a Subject and a Subject Category as separate columns. I would appreciate some help in making this work and also if you had some ideas on how this kind of a problem can be better approached.

import random
import pandas as pd

data=[['Agricultural services', 'Agricultural services, inputs, tools and equipment',   1],
['Agriculture primary production(livestock)',   'Agricultural services, inputs, tools and equipment',7],
['Assist Uganda in upgrading its coffee and cocoa value chains',    'Agricultural services, inputs, tools and equipment',1],
['Building materials and agricultural tools Building and civil works', 'carpentry, construction materials, maintenance, renovation, road works',1],
['Clearing and forwarding services','Clearing and forwarding services', 1],
['Collection of revenue from big slaughters Collection of fees', 'taxes and revenue', 1],
['Collection of revenue from chicken sellers','Collection of fees, taxes and revenue',  19]]
tender_subject = pd.DataFrame(data, columns = ['sub', 'sub_category','subject_dist']) 

subject_tuple=list(tender_subject[['sub', 'sub_category']].itertuples(index=False, name=None)) #we could have also used tuple here instead of 'list'
subject_weights=tender_subject['subject_dist'].tolist()
data={'SL':[], 'Subject':[],'subject_category':[],}
output_df=pd.DataFrame(data)
x=0
for i in range(10):
    p=random.choices(subject_tuple,subject_weights)  
    p1,p2=zip(p) # This line is not working

    output_df.loc[x]=[i]+[p1]+[p2]
    x=x+1
print(output_df.head)

Answer 1

Pandas has rich API, with many methods available for common data processing tasks. You should use these methods where available, because they're thoughtfully designed, robust, and well tested.

Here, you should use the sample method :

tender_subject.sample(10, replace=True, weights=tender_subject.subject_dist)

                                                 sub                                       sub_category  subject_dist
6         Collection of revenue from chicken sellers              Collection of fees, taxes and revenue            19
6         Collection of revenue from chicken sellers              Collection of fees, taxes and revenue            19
6         Collection of revenue from chicken sellers              Collection of fees, taxes and revenue            19
6         Collection of revenue from chicken sellers              Collection of fees, taxes and revenue            19
5  Collection of revenue from big slaughters Coll...                                  taxes and revenue             1
...

You can see that the sampling has resulted in repeated index values. Reset the index and select only the columns you want:

(tender_subject
 .sample(10, replace=True, weights=tender_subject.subject_dist)
 .reset_index()[['sub', 'sub_category']])

                                          sub                                       sub_category
0  Collection of revenue from chicken sellers              Collection of fees, taxes and revenue
1  Collection of revenue from chicken sellers              Collection of fees, taxes and revenue
2  Collection of revenue from chicken sellers              Collection of fees, taxes and revenue
3  Collection of revenue from chicken sellers              Collection of fees, taxes and revenue
4  Collection of revenue from big slaughters               Coll...                                  taxes and revenue
...