I have a data frame with following structure:
df.columns
Index(['first_post_date', 'followers_count', 'friends_count',
'last_post_date','min_retweet', 'retweet_count', 'screen_name',
'tweet_count', 'tweet_with_max_retweet', 'tweets', 'uid'],
dtype='object')
Inside the tweets series, each cell is another data frame containing all the tweets of an user.
df.tweets[0].columns
Index(['created_at', 'id', 'retweet_count', 'text'], dtype='object')
I want to convert this data frame to a multi-index frame, essentially by breaking the cell containing tweets. One index will be the uid, and another will be the id inside tweet.
How can I do that?
So from df, you have tweets columns which contain df of tweets, so I create a tweets_df
dataframe and concat all the df in tweets to tweets_df
, add uid column to know which uid that tweet belong to, then merge info of uid to tweets_df
for further process if needed. Please comment if you need further modification. It is hard to get your sample data and convert to json. So I did this on guessing, hope it still gives you some ideas.
import pandas as pd
df = .... #your df
tweets_df = pd.DataFrame() #create blank df to contain tweets
# explode tweets to df
## loop each uid
for uid in df['uid']:
temp = df.loc[df['uid']==uid, :] # select df by uid
temp = temp['tweets'].iloc[0] # select tweets column -> df
temp['uid'] = uid # add uid column to know tweets belong to which uid
tweets_df = pd.concat([results, temp], ignore_index=True) # concat to container df
# get a uid info df from starting df
uid_info_column = df.columns
uid_info_column.remove('tweets')
uid_info_df = df.loc[:, uid_info_column]
# merge info on uid with tweets_df
final = pd.merge(left=tweets_df, right=uid_info_df, on='uid', how='outer')