如何在Scikit-Learn中重用LabelBinarizer进行输入预测 [英] How to re-use LabelBinarizer for input prediction in Scikit-Learn
问题描述
我使用Scikit-Learn训练了分类器.我正在加载输入以从CSV训练我的分类器.我的某些列(例如镇")的值是规范的(例如可以是"New York","Paris","Stockholm"等).为了使用这些规范列,我正在使用Scikit-Learn的LabelBinarizer进行一次热编码.
I trained a classifier using Scikit-Learn. I am loading the input to train my classifier from a CSV. The value of some of my columns (e.g. 'Town') are canonical (e.g. can be 'New York', 'Paris', 'Stockholm', ...) . In order to use those canonical columns, I am doing one-hot encoding with the LabelBinarizer from Scikit-Learn.
这是我在训练之前如何转换数据的方法:
This is how I transform data before training:
import pandas as pd
from sklearn.preprocessing import LabelBinarizer
headers = [
'Ref.', 'Town' #,...
]
df = pd.read_csv("/path/to/some.csv", header=None, names=headers, na_values="?")
lb = LabelBinarizer()
lb_results = lb.fit_transform(df['Town'])
但是,我不清楚如何使用LabelBinarizer使用要对其进行预测的新输入数据来创建特征向量.特别是,如果新数据包含可见的城镇(例如纽约),则需要在训练数据中将其编码为与同一城镇相同的位置.
It is however not clear to me how to use the LabelBinarizer to create feature vectors using new input data for which I want to do predictions. Especially, if new data contains a seen town (eg New York) it needs to be encoded at the same place as the same town in the training data.
标签二值化应该如何重新应用于新的输入数据?
(如果有人知道如何使用Pandas的get_dummies方法也可以,我对Scikit-Learn的感觉并不强烈.)
推荐答案
对于已经训练好的lb
模型,只需使用lb.transform()
.
Just use lb.transform()
for already trained lb
model.
演示:
假设我们有以下火车DF:
Assuming we have the following train DF:
In [250]: df
Out[250]:
Town
0 New York
1 Munich
2 Kiev
3 Paris
4 Berlin
5 New York
6 Zaporizhzhia
适合(火车)和一步转换(二值化):
Fit (train) & transform (binarize) in one step:
In [251]: r1 = pd.DataFrame(lb.fit_transform(df['Town']), columns=lb.classes_)
收益:
In [252]: r1
Out[252]:
Berlin Kiev Munich New York Paris Zaporizhzhia
0 0 0 0 1 0 0
1 0 0 1 0 0 0
2 0 1 0 0 0 0
3 0 0 0 0 1 0
4 1 0 0 0 0 0
5 0 0 0 1 0 0
6 0 0 0 0 0 1
lb
现在已针对我们在df
Now we can binarize new data sets using trained lb
model (using lb.transform()):
In [253]: new
Out[253]:
Town
0 Munich
1 New York
2 Dubai # <--- new (not trained) town
In [254]: r2 = pd.DataFrame(lb.transform(new['Town']), columns=lb.classes_)
In [255]: r2
Out[255]:
Berlin Kiev Munich New York Paris Zaporizhzhia
0 0 0 1 0 0 0
1 0 0 0 1 0 0
2 0 0 0 0 0 0
这篇关于如何在Scikit-Learn中重用LabelBinarizer进行输入预测的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!