Scikit-Learn Pipeline

A minimal machine learning pipeline using Python and sklearn

Below is a minimal (yet complete) example of a machine learning pipeline using python and scikit-learn and the Palmer Penguins dataset.

Note that the goal here isn’t necessarily to fit the best model; rather it’s just to demonstrate an sklearn pipeline. Also note that I wouldn’t call myself an expert python programmer, so there may be better/more efficient ways to do this.

import polars as pl
import numpy as np
import math
from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler, OneHotEncoder
from sklearn.impute import SimpleImputer
from sklearn.linear_model import LinearRegression
from sklearn.compose import ColumnTransformer
from sklearn.metrics import mean_squared_error

penguins = pl.read_csv(
    "https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2020/2020-07-28/penguins.csv",
    null_values="NA",
)

# filtering to only rows with available body mass data
penguins_complete = penguins.filter(pl.col("body_mass_g").is_not_null())

# coercing null to nan
penguins_complete = penguins_complete.with_columns(pl.all().fill_null(np.nan))

# separate into X and y
y = penguins_complete["body_mass_g"]
X = penguins_complete.select(pl.exclude("body_mass_g"))

# train test split
X_trn, X_tst, y_trn, y_tst = train_test_split(X, y, random_state=408)

# create pipeline for categorical features
cat_feats = ["species", "island", "sex"]
cat_transform = Pipeline(
    [
        ("cat_imputer", SimpleImputer(strategy="most_frequent")),
        ("oh_encoder", OneHotEncoder(drop="first")),
    ]
)

# create pipeline for numerical features
cont_feats = ["bill_length_mm", "bill_depth_mm", "flipper_length_mm", "year"]
cont_transform = Pipeline(
    [
        ("cont_imputer", SimpleImputer(strategy="mean")),
        ("standardizer", StandardScaler()),
    ]
)

# create a preprocessing pipeline
preprocessor = ColumnTransformer(
    [("cats", cat_transform, cat_feats), ("conts", cont_transform, cont_feats)]
)

# make full pipeline
pipe = Pipeline([("preprocess", preprocessor), ("lin_reg", LinearRegression())])

# fit pipeline
pipe.fit(X_trn, y_trn)

# predict training y's
y_hat = pipe.predict(X_trn)

# evaluate model
y_hat_tst = pipe.predict(X_tst)

math.sqrt(mean_squared_error(y_tst, y_hat_tst))

295.13330532494354