Pipeline + cross-validation

Level 4 · Challenge 4

Starter

Prompt

Extend Challenge 3:

1. Add position_group as a categorical feature.
2. Wrap the preprocessing + model in a single Pipeline using ColumnTransformer.
3. Run 5-fold cross-validation.
4. Compare two final models — LinearRegression and RandomForestRegressor(n_estimators=200).

Report mean R² ± std for both. Comment on whether the extra complexity of the RF is justified by the gain.

Expected output

Linear       R² = 0.0XX ± 0.0XX
RandomForest R² = 0.0XX ± 0.0XX

Plus your read on whether the RF earns its complexity.

Hint

from sklearn.compose import ColumnTransformer
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler, OneHotEncoder

preprocess = ColumnTransformer([
    ('num', StandardScaler(), ['prev_yards', 'prev_targets']),
    ('cat', OneHotEncoder(handle_unknown='ignore'), ['position_group']),
])
pipe = Pipeline([('prep', preprocess), ('model', LinearRegression())])

Same SQL as Challenge 3, just expand the SELECT to include position_group and broaden the position filter to IN ('WR', 'TE').

Solution

import pandas as pd
from sklearn.compose import ColumnTransformer
from sklearn.ensemble import RandomForestRegressor
from sklearn.linear_model import LinearRegression
from sklearn.model_selection import cross_val_score
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import OneHotEncoder, StandardScaler
from sqlalchemy import create_engine

eng = create_engine("postgresql+psycopg://onepride:lions@localhost:5432/onepride")

df = pd.read_sql(
    """
    SELECT
        player_display_name, recent_team, season, week,
        receiving_yards, position_group,
        LAG(receiving_yards) OVER w AS prev_yards,
        LAG(targets)         OVER w AS prev_targets
    FROM weekly_stats
    WHERE season BETWEEN 2021 AND 2024
      AND season_type = 'REG'
      AND position_group IN ('WR', 'TE')
      AND targets > 0
    WINDOW w AS (PARTITION BY player_display_name, season ORDER BY week)
    """,
    eng,
)
df = df.dropna(subset=['prev_yards', 'prev_targets'])

X = df[['prev_yards', 'prev_targets', 'position_group']]
y = df['receiving_yards']

preprocess = ColumnTransformer([
    ('num', StandardScaler(), ['prev_yards', 'prev_targets']),
    ('cat', OneHotEncoder(handle_unknown='ignore'), ['position_group']),
])

for name, est in [
    ('Linear',       LinearRegression()),
    ('RandomForest', RandomForestRegressor(n_estimators=200, random_state=42)),
]:
    pipe = Pipeline([('prep', preprocess), ('model', est)])
    scores = cross_val_score(pipe, X, y, cv=5, scoring='r2')
    print(f"{name:13s} R² = {scores.mean():.3f} ± {scores.std():.3f}")

If RF beats Linear by less than one standard deviation, the simpler model wins — fewer hyperparameters, faster, more interpretable.