使用 CountVectorizer 一次对两个 pandas 列进行矢量化

Question

我想一次在两列上应用 Sklearn 的 CountVectorizer。 我试过这个：

features = df[['col 1', 'col2']]
results = df[['col 3']

vectorizer = CountVectorizer(lowercase=False)

features = vectorizer.fit_transform(features)
results = vectorizer.fit_transform(results)

但我收到此错误：

TypeError: expected string or bytes-like object

然后我尝试了这个：

from sklearn.compose import make_column_transformer

vectorizer = CountVectorizer(lowercase=False)
transformer = make_column_transformer((vectorizer, 'col 1'), (vectorizer, 'col 2'))

features = transformer.fit_transform(features)
results = vectorizer.fit_transform(results)

但我收到此错误：

ValueError: Specifying the columns using strings is only supported for pandas DataFrames

我做错了什么，我在这里看到了第二个解决方案：

https://media-exp1.licdn.com/dms/image/C4E22AQFC6Uf5_el2nQ/feedshare-shrink_800/0?e=1591228800&v=beta&t=7ZQbbIvgpQKlTfg1Z_IpGT9DB21LUqy_bkKaNE41l0E

Answer 1

这是解决方案：

features = df.iloc[:, [-2,-3]]
results = df.iloc[:, -1]

from sklearn.compose import make_column_transformer

vectorizer = CountVectorizer(lowercase=False)
transformer = make_column_transformer((vectorizer, 'col 1'), (vectorizer, 'col 2'))

features = transformer.fit_transform(features)
results = vectorizer.fit_transform(results)