Null-Hypothesis Null-Hypothesis - 25 days ago 8
Python Question

Pandas: Add a new column in a data frame based on a value in another data frame

I have two data frames one with a userId, and gender and another data frame that has online activities of these users.

First Data Frame

(df1)


userId, gender
001, F
002, M
003, F
004, M
005, M
006, M


Second data frame
(df2)


userId, itemClicked, ItemBought, date
001, 123182, 123212, 02/02/2016
003, 234256, 123182, 05/02/2016
005, 986834, 234256, 04/19/2016
004, 787663, 787663, 05/12/2016
020, 465738, 465738, 03/20/2016
004, 787223, 787663, 07/12/2016


I want to add gender column to the second data frame by looking up the first data frame based on the userId. df2 might have multiple rows per user since its a click data where same user may have click multiple items.

This is very easy to do in MySql but I am trying to figure out to do it in pandas.

for index, row in df2.iterrows():
user_id = row['userId']
if user_id in df1['userId']:
t = df1.loc[df1['userId'] == user_id]
pdb.set_trace()


Is this the pandas way to so such a task?

Answer
print (df1)
   userId gender
0       1      F
1       2      M
2       3      F
3       4      M
4       5      M
5       6      M

print (df2)
   userId  itemClicked  ItemBought        date
0       1       123182      123212  02/02/2016
1       3       234256      123182  05/02/2016
2       5       986834      234256  04/19/2016
3       4       787663      787663  05/12/2016
4      20       465738      465738  03/20/2016
5       4       787223      787663  07/12/2016

You can use map:

df2['gender'] = df2.userId.map(df1.set_index('userId')['gender'].to_dict())

print (df2)
   userId  itemClicked  ItemBought        date gender
0       1       123182      123212  02/02/2016      F
1       3       234256      123182  05/02/2016      F
2       5       986834      234256  04/19/2016      M
3       4       787663      787663  05/12/2016      M
4      20       465738      465738  03/20/2016    NaN
5       4       787223      787663  07/12/2016      M

Another solution with merge and left join, parameter on can be omit if only column gender is same in both DataFrames:

df = pd.merge(df2, df1, how='left')

print (df)
   userId  itemClicked  ItemBought        date gender
0       1       123182      123212  02/02/2016      F
1       3       234256      123182  05/02/2016      F
2       5       986834      234256  04/19/2016      M
3       4       787663      787663  05/12/2016      M
4      20       465738      465738  03/20/2016    NaN
5       4       787223      787663  07/12/2016      M

Timings:

#len(df2) = 600k
df2 = pd.concat([df2]*100000).reset_index(drop=True)

def f(df1,df2):
    df2['gender'] = df2.userId.map(df1.set_index('userId')['gender'].to_dict())
    return df2


In [43]: %timeit f(df1,df2)
10 loops, best of 3: 34.2 ms per loop

In [44]: %timeit (pd.merge(df2, df1, how='left'))
10 loops, best of 3: 102 ms per loop