pandas apply typeError: 'float' object is not subscriptable

Question

I have a dataframe df_tr like this:

      item_id    target   target_sum  target_count
0        0          0           1            50            
1        0          0           1            50

I'm trying to find the mean of the target but excluding the target value of the current row, and put the mean value in a new column. The result would be:

     item_id    target   target_sum  target_count item_id_mean_target
0        0          0           1            50           0.02041
1        0          0           1            50           0.02041

where I computed item_id_mean_target value from the formula:

target_sum - target/target_count - 1

...with this code:

df_tr['item_id_mean_target'] = df_tr.target.apply(lambda x: (x['target_sum']-x)/(x['target_count']-1))

I think my solution is correct but instead I got:

TypeError: 'float' object is not subscriptable

score 4 · Accepted Answer · answered Jan 15 '18 at 06:02

No need for apply here, pandas (and therefore numpy) broadcasts operations.

df['item_id_mean_target'] = (df.target_sum - df.target) / (df.target_count - 1)

df

   item_id  target  target_sum  target_count  item_id_mean_target
0        0       0           1            50             0.020408
1        0       0           1            50             0.020408

As for why your error occurs, you are calling apply on a pd.Series object, therefore, you cannot reference any other columns inside the apply (since it only receives scalar values).

To fix it, you'd need to do df.apply(...) but at that point, you're penalised with low performance, so, I wouldn't recommend doing it.

Alexander · Answer 2 · 2018-01-15T06:06:52.290

1

Ignoring the sum and count columns and using groupby to derive them:

df_tr.groupby('item_id').apply(lambda x: (x['target'].sum() - x['target'])     
                                         / (x['target'].count() - 1))

You may also notice the issue in your original statement where you had x['target_sum']-x. It should have been x['target_sum']-x['target'].

edited Jan 15 '18 at 06:06

answered Jan 15 '18 at 06:04

Alexander

105,104
32
201
196

Smart, you've gone beyond the problem. – cs95 Jan 15 '18 at 06:05

score 0 · Answer 3 · answered Jan 15 '18 at 06:01

0

try this:

df_tr.apply(lambda x:(x['target_sum']-x)/(x['target_count']-1), axis=1)

answered Jan 15 '18 at 06:01

Advay Umare

432
10
23

This works, but is slow compared to a purely vectorised operation. – cs95 Jan 15 '18 at 06:03
Thank you for this code snippet, which might provide some limited, immediate help. A [proper explanation would greatly improve its long-term value](//meta.stackexchange.com/q/114762/350567) by showing *why* this is a good solution to the problem, and would make it more useful to future readers with other, similar questions. Please [edit] your answer to add some explanation, including the assumptions you've made. – iBug Jan 15 '18 at 08:02

pandas apply typeError: 'float' object is not subscriptable

3 Answers3