如何修改pd.dataframe中的列值

Question

Background: Actually I wanna modify the value in dataframe, only top 20 sport should be kept, and the others should be displayed like "Others". It's a copy of existed columns, as following:

athlete_events['Sport_modified'] = athlete_events['Sport']

And the filter that contains top20 sport name is generated like:

top20_sport = athlete_events['Sport'].value_counts().head(20).index

And the modify process is like following: Method 1:

 def classify_sports(cols, filters):
for i in cols:
    if i in filters:
        pass
    else:
        i = 'Others'
classify_sports(athlete_events.Sport_modified, top20_sport)

Method 2:

athlete_events.Sport_modified.apply(lambda x : x if x in top20_sport else 'Others')

However, the 2 method above did not works. The only way I could do just like this code:

athlete_events.loc[
(athlete_events['Sport'] !='Athletics')&
(athlete_events['Sport'] !='Gymnastics')&
(athlete_events['Sport'] !='Swimming')&
(athlete_events['Sport'] !='Shooting')&
(athlete_events['Sport'] !='Cycling')&
(athlete_events['Sport'] !='Fencing')&
(athlete_events['Sport'] !='Rowing')&
(athlete_events['Sport'] !='Cross Country Skiing')&
(athlete_events['Sport'] !='Alpine Skiing')&
(athlete_events['Sport'] !='Wrestling')&
(athlete_events['Sport'] !='Football')&
(athlete_events['Sport'] !='Sailing')&
(athlete_events['Sport'] !='Equestrianism')&
(athlete_events['Sport'] !='Canoeing')&
(athlete_events['Sport'] !='Boxing')&
(athlete_events['Sport'] !='Speed Skating')&
(athlete_events['Sport'] !='Ice Hockey')&
(athlete_events['Sport'] !='Hockey')&
(athlete_events['Sport'] !='Biathlon')&
(athlete_events['Sport'] !='Basketball')
,'Sport_modified'] = 'Others'

What's the problems of that 2 ways above? Thanks for help.

Answer 1

Your first method will never work, since your function does not return a series, nor does it return anything for a row-wise calculation.

Your second method is not in-place , you need to assign back to a series. For instance:

df['sport_modified'] = df['sport'].apply(lambda x : x if x in top20_sport else 'Others')

Your final solution can be more efficiently expressed using pd.Series.isin , negated via ~ :

L = ['Athletics', 'Gymnastics', ...]

df.loc[~df['sport'].isin(L), 'sport_modified'] = 'Others'