Add multiple values in a row of a dataframe while iterate trough a list

user_13551189 0
Viewed 84

I am building a recommender system of foods and I have a dataframe:

df:
            meat vegetables cheese ketchup egg...
hamburger     3      5        2       2     1   
    pasta     0      0        4       0     1    
     soup     0      2        0       0     0     
      ...

I also have a list which contains ingredients that an user does not like:

dislike:["cheese", "egg"]

So what I am trying to do is to create a function which adds a new row "user_name" with a 10 in those ingredients that he/she does not like and a 0 in all the others columns. Output should be:

            meat vegetables cheese ketchup egg...
hamburger     3      5        2       2     1   
    pasta     0      0        4       0     1    
     soup     0      2        0       0     0     
 new_user     0      0       10       0    10
...

I have simplify the dataframe and the list in order to make it more comprehensive, but they are actually way more longer.

This is what I have write until now:

def user_pre(df):
    dislike=["cheese","egg"]
    for ing in dislike:
            df.loc["new_user"]= pd.Series({ing:10})
    return df

I "works" but only for the last element in dislike list. Besides it does not add a 0 in the other cells but a Nan.

Thank you so much in advance!

pandas dataframe for-loop list python
4 Answers

I am not sure how "healthy" it is to mix users with dishes in a single pandas DataFrame but a function like this should do the work:

def insert_user_dislikes(user_name='new_user', df=df, ingredients=['meat', 'egg']):
    df.loc[user_name] = [10 if col in ingredients else 0 for col in df.columns]

insert_user_dislikes('new_user', df, ['meat', 'egg'])

Edit 1: I like @Fred's Solution as well:

def insert_user_dislikes2(user_name='new_user', df=df, ingredients=['meat', 'egg']):
    df.loc[user_name] = 0
    df.loc[user_name, ingredients] = 10
insert_user_dislikes('user_name', df, ['meat', 'egg'])

Edit 2: Here is Shubham's solution for performance assessment:

def insert_user_dislikes3(user_name='new_user', df=df, ingredients=['meat', 'egg']):
    s = pd.Series(
        np.where(df.columns.isin(ingredients), 10, 0), 
        name=user_name, index=df.columns, dtype='int')
    return df.append(s)

In term of performance (on a very small dataset), it looks like the list comprehension one is faster though:

df = pd.DataFrame([[3, 5, 2, 2, 1],
   [0, 0, 4, 0, 1]],
   columns=['meat', 'vegetables', 'cheese','ketchup', 'egg'],
   index=['hamburger', 'pasta'])

print(timeit.timeit(insert_user_dislikes, number=1000))
0.125

print(timeit.timeit(insert_user_dislikes2, number=1000))
0.547

print(timeit.timeit(insert_user_dislikes3, number=1000))
2.153
user_117795930

I'm not sure about how efficient the approach is, but this should work

dislikes = ["cheese","egg"]
new_user = "Tom"
df.loc[new_user] = 0
for dislike in dislikes:
    if dislike not in df.columns:
        df[dislike] = 0
    df.loc[new_user, dislike] = 10
user_80502310

Set the new_user row = to zero, then filter and equal to 10. 

print(df)
          meat  vegetables  cheese  ketchup  egg
hamburger     3           5       2        2    1
pasta         0           0       4        0    1
soup          0           2       0        0    0

Create new_user as zero.

df.loc["new_user", :] = 0
print(df)
          meat  vegetables  cheese  ketchup  egg
hamburger   3.0         5.0     2.0      2.0  1.0
pasta       0.0         0.0     4.0      0.0  1.0
soup        0.0         2.0     0.0      0.0  0.0
new_user    0.0         0.0     0.0      0.0  0.0

Then again but filtered and set to 10.

dislike = ["cheese", "egg"]

df.loc["new_user", dislike] = 10
print(df)
           meat  vegetables  cheese  ketchup   egg
hamburger   3.0         5.0     2.0      2.0   1.0
pasta       0.0         0.0     4.0      0.0   1.0
soup        0.0         2.0     0.0      0.0   0.0
new_user    0.0         0.0    10.0      0.0  10.0
user_108886550

You can use Series.isin to check which column values of dataframe are present in dislike list, then you can use DataFrame.append to append the newly created series s to the original dataframe df.

Use:

import numpy as np

s = pd.Series(
    np.where(df.columns.isin(dislike), 10, 0), 
    name='new_user', index=df.columns, dtype='int') # create a new pandas series

df = df.append(s)

The resulting dataframe df will be:

           meat  vegetables  cheese  ketchup  egg                                            
hamburger     3           5       2        2    1
pasta         0           0       4        0    1
soup          0           2       0        0    0
new_user      0           0      10        0   10
user_128331660
Related