I have a pandas dataframe which has a 'source' column as shown in the table below. I want to normalize its values and create a new column called 'derived'.

source	value	derived
google	1	google
Google	1	google
googlechannel	2	google
facebook	2	facebook
Facebook	2	facebook
lt_Facebook	4	facebook
twitter	9	other
snapchat	10	other

I tried the code but it is giving me error.

import pandas as pd
data_dict = {'source':['google','Google','googlechannel','facebook','Facebook','lt_Facebook','twitter','snapchat'],
       'value':[1,1,2,2,2,4,9,10]}
data = pd.DataFrame.from_dict(data_dict)

def normalize_source(data):
    if data['source'].str.lower().str.contains('facebook'):
        return 'facebook'
    elif data['source'].lower().str.contains('google'):
        return 'google'
    else:
        return 'other'

data.loc[:,'derived'] = data.apply(normalize_source,axis=1)   
data.head()

I am getting the following error:

AttributeError: 'str' object has no attribute 'str'

CodePudding user response：

you need to make use of __ contains __ here is your updated code

data_dict = {'source':['google','Google','googlechannel','facebook','Facebook','lt_Facebook','twitter','snapchat'],
       'value':[1,1,2,2,2,4,9,10]}
data = pd.DataFrame.from_dict(data_dict)

def normalize_source(data):
    if data['source'].lower().__contains__('facebook'):
        return 'facebook'
    elif data['source'].lower().__contains__('google'):
        return 'google'
    else:
        return 'other'


data.loc[:,'derived'] = data.apply(normalize_source,axis=1)   
data.head()

    source      value   derived
0   google          1   google
1   Google          1   google
2   googlechannel   2   google
3   facebook        2   facebook
4   Facebook        2   facebook

CodePudding user response：

You are trying to apply a function which, as coded, takes a whole DataFrame.

Here is a fix:

def normalize_source(x):
    x = x.lower()
    if 'facebook' in x:
        return 'facebook'
    elif 'google' in x:
        return 'google'
    return 'other'

data = data.assign(derived=data['source'].apply(normalize_source))

>>> data
          source  value   derived
0         google      1    google
1         Google      1    google
2  googlechannel      2    google
3       facebook      2  facebook
4       Facebook      2  facebook
5    lt_Facebook      4  facebook
6        twitter      9     other
7       snapchat     10     other

Alternative (for the specific problem at hand):

data = data.assign(
    derived=data['source']
    .str.lower()
    .str.extract(r'(google|facebook)')
    .fillna('other')
)

Depending on the number of matching possibilities, this could be faster (as regex builds an optimized parser).

CodePudding user response：

How about a conditional replacement via regular expression? In particular,

companyList = ['facebook', 'google', 'twitter', 'snapchat']

for compName in companyList:
    data['source'] = data['source'].str.lower().replace(r'(^.*{}.*$)'.format(compName), compName, regex=True)

yields

data