How to select multiple (numerical & text) columns using sklearn Pipeline & FeatureUnion for text classification?

Question

I have developed a text model for multilabel classification. The OneVsRestClassifier LinearSVC model uses sklearns Pipeline and FeatureUnion for model preparation.

The primary input features consist of a text column called response but also 5 topic probabilities (generated from a previous LDA Topic Model) called t1_prob - t5_prob to predict the 5 possible labels. There are other feature creation steps in the pipeline for the the generation of the TfidfVectorizer .

I ended up calling each column with ItemSelector and performing the ArrayCaster (see the code below for function definition) 5 times on these topic probability columns individually. Is there a better way to use FeatureUnion to select multiple columns in a pipeline? (so I don't have to do it 5 times)

I am wondering if it is necessary to duplicate the topic1_feature - topic5_feature code or if multiple columns can be selected in a more concise way?

The data I am feeding in is a Pandas dataFrame:

id response label_1 label_2 label3  label_4 label_5     t1_prob t2_prob t3_prob t4_prob t5_prob
1   Text from response...   0.0 0.0 0.0 0.0 0.0 0.0     0.0625  0.0625  0.1875  0.0625  0.1250
2   Text to model with...   0.0 0.0 0.0 0.0 0.0 0.0     0.1333  0.1333  0.0667  0.0667  0.0667  
3   Text to work with ...   0.0 0.0 0.0 0.0 0.0 0.0     0.1111  0.0938  0.0393  0.0198  0.2759  
4   Free text comments ...  0.0 0.0 1.0 1.0 0.0 0.0     0.2162  0.1104  0.0341  0.0847  0.0559  

The x_train is response and the 5 topic probability columns (t1_prob, t2_prob, t3_prob, t4_prob, t5_prob).

The y_train is the 5 label columns which I have called .values on to return a numpy representation of the DataFrame. (label_1, label_2, label3, label_4, label_5)

Sample DataFrame:

import pandas as pd
column_headers = ["id", "response", 
                  "label_1", "label_2", "label3", "label_4", "label_5",
                  "t1_prob", "t2_prob", "t3_prob", "t4_prob", "t5_prob"]

input_data = [
    [1, "Text from response",0.0,0.0,1.0,0.0,0.0,0.0625,0.0625,0.1875,0.0625,0.1250],
    [2, "Text to model with",0.0,0.0,0.0,0.0,0.0,0.1333,0.1333,0.0667,0.0667,0.0667],
    [3, "Text to work with",0.0,0.0,0.0,0.0,0.0,0.1111,0.0938,0.0393,0.0198,0.2759],
    [4, "Free text comments",0.0,0.0,1.0,1.0,1.0,0.2162,0.1104,0.0341,0.0847,0.0559]
    ]

df = pd.DataFrame(input_data, columns = column_headers)
df = df.set_index('id')
df

I think my implementation is a little bit round about because FeatureUnion will only handle 2-D arrays when combining them, so any other type like DataFrame have been problematic for me. However, this example works--I am just looking for ways to improve it and make it more DRY.

from sklearn.pipeline import Pipeline, FeatureUnion
from sklearn.base import BaseEstimator, TransformerMixin

class ItemSelector(BaseEstimator, TransformerMixin):
    def __init__(self, column):
        self.column = column

    def fit(self, X, y=None):
        return self

    def transform(self, X, y=None):
        return X[self.column]

class ArrayCaster(BaseEstimator, TransformerMixin):
    def fit(self, x, y=None):
        return self

    def transform(self, data):
        return np.transpose(np.matrix(data))


def basic_text_model(trainX, testX, trainY, testY, classLabels, plotPath):
    '''OneVsRestClassifier for multi-label prediction''' 
pipeline = Pipeline([
    ('features', FeatureUnion([
            ('topic1_feature', Pipeline([
                ('selector', ItemSelector(column='t1_prob')),
                ('caster', ArrayCaster())
            ])),
            ('topic2_feature', Pipeline([
                ('selector', ItemSelector(column='t2_prob')),
                ('caster', ArrayCaster())
            ])),
            ('topic3_feature', Pipeline([
                ('selector', ItemSelector(column='t3_prob')),
                ('caster', ArrayCaster())
            ])),
            ('topic4_feature', Pipeline([
                ('selector', ItemSelector(column='t4_prob')),
                ('caster', ArrayCaster())
            ])),
            ('topic5_feature', Pipeline([
                ('selector', ItemSelector(column='t5_prob')),
                ('caster', ArrayCaster())
            ])),
           ('word_features', Pipeline([
                    ('vect', CountVectorizer(analyzer="word", stop_words='english')), 
                    ('tfidf', TfidfTransformer(use_idf = True)),
            ])),
     ])),
    ('clf', OneVsRestClassifier(svm.LinearSVC(random_state=random_state))) 
])

# Fit the model
pipeline.fit(trainX, trainY)
predicted = pipeline.predict(testX)

My incorporation of ArrayCaster into the process arose from this answer .

Answer 1

I figured out the answer to this question using the FunctionTransformer inspired by @Marcus V's solution to this question . The revised pipeline is much more succinct.

from sklearn.preprocessing import FunctionTransformer

get_numeric_data = FunctionTransformer(lambda x: x[['t1_prob', 't2_prob', 't3_prob', 't4_prob', 't5_prob']], validate=False)

pipeline = Pipeline(
    [
        (
            "features",
            FeatureUnion(
                [
                    ("numeric_features", Pipeline([("selector", get_numeric_data)])),
                    (
                        "word_features",
                        Pipeline(
                            [
                                ("vect", CountVectorizer(analyzer="word", stop_words="english")),
                                ("tfidf", TfidfTransformer(use_idf=True)),
                            ]
                        ),
                    ),
                ]
            ),
        ),
        ("clf", OneVsRestClassifier(svm.LinearSVC(random_state=10))),
    ]
)