CODEWITHSUNDEEP
pythonpandasbinning
Aurangzeb Rathore
Aurangzeb Rathore

Reputation: 147

How to bin string values according to list of strings?

Is there a way to bin a pandas column of STRINGS into custom groups of custom names. Like similar to the cuts function but for strings.

For example maybe using a list of lists to define what groups are. 

grouping_lists = [['Pakistan', 'China', 'Iran'], ['Germany', 'UK', 'Poland'], 
                  ['Australia'], ['USA']]

Corresponding to names ['Asia', 'Europe', 'Australia', 'Other'].

And should something not exist in the list, it get's marked as 'Other' or something.

Example:

          my_id  country_name
    0     100     Pakistan
    1     200     Germany
    2     140     Australia
    3     400     Germany
    4     225     China
    5     125     Pakistan
    6     600     Poland
    7       0     Austria


          my_id  country_name  Groups
    0     100     Pakistan      Asia
    1     200     Germany       Europe
    2     140     Australia     Australia
    3     400     Germany       Europe
    4     225     China         Asia
    5     125     Pakistan      Asia 
    6     600     Poland        Europe
    7       0     Austria       Other

Upvotes: 1

Views: 2698

Answers (3)

eva-vw
eva-vw

Reputation: 670

You can use a lambda if you're not worried about speed.

groups = {
    "Asia": ["Pakistan", "China", "Iran"],
    "Europe": ["Germany", "UK", "Poland"],
    "Australia": ["Australia"],
}
df["Groups"] = (
    df["country_names"]
    .apply(lambda x: [k for k in groups.keys() if x in groups[k]])
    .str[0]
    .fillna("Other")
)

Upvotes: 1

YOLO
YOLO

Reputation: 21749

Here's a way where you can don't need to create map dict manually (in case it is big):

grouping_lists = [['Pakistan', 'China', 'Iran'], ['Germany', 'UK', 'Poland'], 
                  ['Australia'], ['USA']] 
names =  ['Asia', 'Europe', 'Australia', 'Other']

# create a df with mapping information
maps = (pd.DataFrame({'Groups': names, 'country_name': grouping_lists})
        .explode('country_name')
        .reset_index(drop=True))

# join maps
df = df.merge(maps, on = 'country_name', how='left').fillna("Other")

   my_id country_name     Groups
0    100     Pakistan       Asia
1    200      Germany     Europe
2    140    Australia  Australia
3    400      Germany     Europe
4    225        China       Asia
5    125     Pakistan       Asia
6    600       Poland     Europe
7      0      Austria      Other

Upvotes: 3

gold_cy
gold_cy

Reputation: 14236

Instead of binning your answers you could modify your grouping lists to be a dictionary and then use pandas.Series.map

country_map = {
    'Pakistan': 'Asia', 'China': 'Asia', 
    'Iran': 'Asia', 'Germany': 'Europe', 
    'UK': 'Europe', 'Poland': 'Europe', 
    'Australia': 'Australia', 'USA': 'Other'
}

df.assign(Groups=df.country_name.map(country_map)).fillna('Other')

   my_id country_name     Groups
0    100     Pakistan       Asia
1    200      Germany     Europe
2    140    Australia  Australia
3    400      Germany     Europe
4    225        China       Asia
5    125     Pakistan       Asia
6    600       Poland     Europe
7      0      Austria      Other

Upvotes: 1

Related Questions