Python pandas create multiple columns based on unique values of one column-CodePudding

I have a pandas DataFrame,

account_id, campaign_objective, campaign_spend, conversions
__________, __________________, ______________, ___________
1,          sales,              100,            25
1,          brand,              50,             25
2,          sales,              80,             12
2,          brand,              60,             12

What I would like to do is create one column for each unique campaign_objective and assign it the corresponding spend value.

account_id, sales, brand, conversions
__________, _____, _____, ___________
1,          100,   50,    25
2,          80,    60,    12

My approach has been using for loops and dictionaries. This is suboptimal as I have 20 million rows in my dataframe and there are 100 campaign_objectives; in other words, my for loop will need to iterate through 2 billion values.

new_df = {'account_id':[], 'conversions':[]}
for obj in obj_goal_list:
    new_df.update({obj:[]})

for acct in df['account_id'].unique():
    acct_df = df[df['account_id']==acct]
    new_df['account_id'].append(acct)
    new_df['conversions'].append(acct_df['conversions'])

    for obj in obj_goal_list: 
        if obj in acct_df['objective_and_goal']:
            spend = acct_df[acct_df['objective_and_goal']==obj]['spend']
            new_df[obj].append(spend)
        else:
            new_df[obj].append(0.0)
new_df = pd.DataFrame(new_df)

I'm curious if there's a more "Pandas" way to accomplish this by pivots or other means?

CodePudding user response：

here is one way to do it

df.pivot(index=['account_id','conversions'], columns='campaign_objective', values='campaign_spend')

     campaign_objective     brand   sales
account_id  conversions         
         1           25        50     100
         2           12        60      80

with reset_index

df.pivot(index=['account_id','conversions'], columns='campaign_objective', values='campaign_spend').reset_index()

campaign_objective  account_id  conversions     brand   sales
                 0           1           25        50     100
                 1           2           12        60      80

CodePudding user response：

Pivot method may help you.

df.pivot(index = ['account_id', 'conversions'], columns = 'campaign_objective', values = 'campaign_spend').reset_index()