获取包含字符串列表的熊猫列的词频

时间:2021-01-22 11:32:57

标签: python pandas dataframe

我有一个熊猫数据框:

import pandas as pd
test = pd.DataFrame({'words':[['foo','bar none','scare','bar','foo'],
                              ['race','bar none','scare'],
                              ['ten','scare','crow bird']]})

我正在尝试获取数据框列中所有列表元素的单词/短语计数。我目前的解决方案是:

allwords = []

for index, row in test.iterrows():
    for word in row['words']:
        allwords.append(word)
from collections import Counter
pd.Series(Counter(allwords)).sort_values(ascending=False)

这有效,但我想知道是否有更快的解决方案。注意:我没有使用 ' '.join(),因为我不希望将短语拆分为单个单词。

3 个答案:

答案 0 :(得分:3)

尝试使用 Counter

import collections
words = test['words'].tolist()

collections.Counter([x for sublist in words for x in sublist])

Counter({'foo': 2,
         'bar none': 2,
         'scare': 3,
         'bar': 1,
         'race': 1,
         'ten': 1,
         'crow bird': 1})

答案 1 :(得分:2)

为了提高性能不要使用 iterrows:

from collections import Counter
from  itertools import chain

a = pd.Series(Counter(chain.from_iterable(test['words']))).sort_values(ascending=False)
print (a)
scare        3
foo          2
bar none     2
bar          1
race         1
ten          1
crow bird    1
dtype: int64

仅熊猫解决方案:

a = pd.Series([y for x in test['words'] for y in x]).value_counts()
print (a)
scare        3
bar none     2
foo          2
bar          1
race         1
crow bird    1
ten          1
dtype: int64

答案 2 :(得分:2)

让我们用 .hstack 试试 .value_counts

pd.value_counts(np.hstack(test['words']))

scare        3
foo          2
bar none     2
ten          1
bar          1
crow bird    1
race         1
dtype: int64