Python: Combining Low Frequency Factors/Category Counts

Question

There is a great solution in R.

My df.column looks like:

Windows
Windows
Mac
Mac
Mac
Linux
Windows
...

I want to replace low frequency categories with 'Other' in this df.column vector. For example, I need my df.column to look like

Windows
Windows
Mac
Mac
Mac
Linux -> Other
Windows
...

I would like to rename these rare categories, to reduce the number of factors in my regression. This is why I need the original vector. In python, after running the command to get the frequency table I get:

pd.value_counts(df.column)


Windows          26083
iOS              19711
Android          13077
Macintosh         5799
Chrome OS          347
Linux              285
Windows Phone      167
(not set)           22
BlackBerry          11

I wonder if there is a method to rename 'Chrome OS', 'Linux' (low frequency data) into another category (for example category 'Other', and do so in an efficient way.

Answer 1

Mask by finding percentage of occupency i.e :

series = pd.value_counts(df.column)
mask = (series/series.sum() * 100).lt(1)
# To replace df['column'] use np.where I.e 
df['column'] = np.where(df['column'].isin(series[mask].index),'Other',df['column'])

To change the index with sum :

new = series[~mask]
new['Other'] = series[mask].sum()

Windows      26083
iOS          19711
Android      13077
Macintosh     5799
Other          832
Name: 1, dtype: int64

If you want to replace the index then :

series.index = np.where(series.index.isin(series[mask].index),'Other',series.index)

Windows      26083
iOS          19711
Android      13077
Macintosh     5799
Other          347
Other          285
Other          167
Other           22
Other           11
Name: 1, dtype: int64

Explanation

(series/series.sum() * 100) # This will give you the percentage i.e 

Windows          39.820158
iOS              30.092211
Android          19.964276
Macintosh         8.853165
Chrome OS         0.529755
Linux             0.435101
Windows Phone     0.254954
(not set)         0.033587
BlackBerry        0.016793
Name: 1, dtype: float64

.lt(1) is equivalent to lesser than 1. That gives you a boolean mask, based on that mask index and assign the data

Answer 2

This is a (late) extension to your question; it applies the rationale of combining low-frequency categories (in proportion less than min_freq) to the columns of an entire dataframe. It is based on @Bharath's answer.

def condense_category(col, min_freq=0.01, new_name='other'):
    series = pd.value_counts(col)
    mask = (series/series.sum()).lt(min_freq)
    return pd.Series(np.where(col.isin(series[mask].index), new_name, col))

A simple example of application:

df_toy = pd.DataFrame({'x': [1, 2, 3, 4] + [5]*100, 'y': [5, 6, 7, 8] + [0]*100})
df_toy = df_toy.apply(condense_category, axis=0)
print(df_toy)

#          x      y
# 0    other  other
# 1    other  other
# 2    other  other
# 3    other  other
# 4        5      0
# ..     ...    ...
# 99       5      0
# 100      5      0
# 101      5      0
# 102      5      0
# 103      5      0
# 
# [104 rows x 2 columns]