Summary - Pipeline

In this group you learned the Pipeline structure and advantages.

Summary: pipelines

In this group:

A pipeline makes preprocessing and the model one estimator; it prevents leakage structurally, especially for steps that look at the target.
Intermediate steps must be transformers; steps are reached with pipe["name"], pipe[:-1]; "passthrough" turns a step off.
fit_transform runs in fit, transform in predict.
With joblib the whole chain is saved and moved to production.
ColumnTransformer applies different transformations to different column groups.

Exercise End-to-end chain

For the mixed table, build a make_pipeline from a ColumnTransformer (numbers with StandardScaler, city with OneHotEncoder(handle_unknown="ignore")) and LogisticRegression(), and train it with y. Print the prediction for a new row containing the city "Antalya", unseen in training.

Show example solution

import numpy as np
import pandas as pd
from sklearn.compose import ColumnTransformer
from sklearn.preprocessing import StandardScaler, OneHotEncoder

df = pd.DataFrame({
    "yas": [25, 32, 47, 51],
    "gelir": [3000, 4500, 8000, 6000],
    "sehir": ["Ankara", "İzmir", "Ankara", "Bursa"],
})
from sklearn.pipeline import make_pipeline
from sklearn.linear_model import LogisticRegression

y = np.array([0, 0, 1, 1])
yeni = pd.DataFrame({"yas": [50], "gelir": [7500], "sehir": ["Antalya"]})

ct = ColumnTransformer([("sayi", StandardScaler(), ["yas", "gelir"]), ("kat", OneHotEncoder(handle_unknown="ignore"), ["sehir"])])
model = make_pipeline(ct, LogisticRegression()).fit(df, y)
print(model.predict(yeni))

Test yourself

How is the <code>gamma</code> parameter of an SVC inside a pipeline written in a grid (step name "svc")?
- "svc_gamma"
- "svc__gamma"
- "gamma"
- "svc.gamma"
Show answer

"svc__gamma" — The step name and the parameter name are separated by a double underscore.

Combine preprocessing and the model with automatic names.

model = ____(ct, LogisticRegression())

Show answer

model = make_pipeline(ct, LogisticRegression())

Match the tools.

Pipeline

Sequential steps
ColumnTransformer

Parallel column groups
joblib

Saving
passthrough

Turning a step off

Next lesson: What is cross-validation?

We Are Enhancing Your Experience

Cookie Preferences

Strictly Necessary Cookies

Analytical Cookies