ezpipe: Data processing pipelines for the people

Building a machine learning pipeline involves exploring different models, pre and post-processing, feature selection and data cleaning, ezpipe makes the process of wiring up all those different steps in an easy, efficient and reproducible way.

Here is an example on how to setup a pipeline for sentiment analysis.

from ezpipe import Pipeline
p = Pipeline()
p.add_map('X', 'tokenize', tokenize)
p.add_transformer('tokenize', 'vectorized', CountVectorizer(), name='vec')
p.add_model('vectorized', 'sentiment', LogisticRegression(), name='reg')

Train the pipeline.

X = ['Good boy',
     'Bad boy']
y = [1, 0]
p.fit('vec', X=X)
p.fit('reg', X=X, sentiment=y)

Make predictions

p.get('sentiment', X=['Good girl', 'Bad girl'])
# [1, 0]

Contents:

Indices and tables