Building Better Models Practice

Practise ensembles, feature engineering, pipelines with mixed columns, cross-validation, hyperparameter tuning and imbalanced data with scikit-learn.

Intermediate7 problemsUpdated October 24, 2026

Review first: Random Forests, Gradient Boosting, Feature Engineering, Pipelines and Column Transformers, Cross-Validation, Hyperparameter Tuning, Imbalanced Data

1.Forest versus single tree

Intermediate

Compare a DecisionTreeClassifier(random_state=0) with a RandomForestClassifier(n_estimators=200, random_state=0) on the breast cancer data using 5-fold cross-validation. Print each model's mean accuracy (3 decimals).

Show hint

cross_val_score(model, X, y, cv=5).mean() for each model.

Show solution

Python

from sklearn.datasets import load_breast_cancer
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import cross_val_score
from sklearn.tree import DecisionTreeClassifier

X, y = load_breast_cancer(return_X_y=True)
for model in [DecisionTreeClassifier(random_state=0), RandomForestClassifier(n_estimators=200, random_state=0)]:
    print(type(model).__name__, round(cross_val_score(model, X, y, cv=5).mean(), 3))

Output

DecisionTreeClassifier 0.917
RandomForestClassifier 0.96

Averaging 200 trees, each trained on different samples and features, reduces variance, so the forest beats the single tree.

2.Most important features

Intermediate

Fit RandomForestClassifier(n_estimators=200, random_state=0) on the full breast cancer data and print the names of the three features with the highest feature_importances_, most important first.

Show hint

Put the importances in a pandas Series indexed by data.feature_names and use nlargest(3).

Show solution

Python

import pandas as pd
from sklearn.datasets import load_breast_cancer
from sklearn.ensemble import RandomForestClassifier

data = load_breast_cancer()
forest = RandomForestClassifier(n_estimators=200, random_state=0).fit(data.data, data.target)
importance = pd.Series(forest.feature_importances_, index=data.feature_names)
print(importance.nlargest(3).index.tolist())

Output

['worst concave points', 'worst perimeter', 'worst radius']

3.Mixed column pipeline

Intermediate

Customers have age, income and city columns. Build a ColumnTransformer that scales the numeric columns and one-hot encodes city, fit it on the data below, and print the output feature names and the shape of the transformed data.

Python

df = pd.DataFrame({"age": [25, 40, 33, 51], "income": [30000, 82000, 54000, 91000],
                   "city": ["Pune", "Delhi", "Pune", "Chennai"]})

Show hint

ColumnTransformer([("num", StandardScaler(), ["age", "income"]), ("cat", OneHotEncoder(), ["city"])]), then get_feature_names_out().

Show solution

Python

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

df = pd.DataFrame({"age": [25, 40, 33, 51], "income": [30000, 82000, 54000, 91000],
                   "city": ["Pune", "Delhi", "Pune", "Chennai"]})
ct = ColumnTransformer([("num", StandardScaler(), ["age", "income"]), ("cat", OneHotEncoder(), ["city"])])
out = ct.fit_transform(df)
print(ct.get_feature_names_out().tolist())
print(out.shape)

Output

['num__age', 'num__income', 'cat__city_Chennai', 'cat__city_Delhi', 'cat__city_Pune']
(4, 5)

4.Features from dates

Intermediate

From the order dates ["2026-01-05", "2026-01-10", "2026-02-18", "2026-03-01"], build a DataFrame with columns month, day_of_week (Monday = 0) and is_weekend (True for Saturday and Sunday), and print it.

Show hint

Convert with pd.to_datetime, then use .dt.month and .dt.dayofweek.

Show solution

Python

import pandas as pd

dates = pd.to_datetime(pd.Series(["2026-01-05", "2026-01-10", "2026-02-18", "2026-03-01"]))
features = pd.DataFrame({"month": dates.dt.month, "day_of_week": dates.dt.dayofweek,
                         "is_weekend": dates.dt.dayofweek >= 5})
print(features)

Output

   month  day_of_week  is_weekend
0      1            0       False
1      1            5        True
2      2            2       False
3      3            6        True

5.Cross-validation spread

Intermediate

Run 10-fold stratified cross-validation (shuffle=True, random_state=0) of a scaled logistic regression on the breast cancer data. Print the lowest, highest and mean fold accuracy (3 decimals).

Show hint

Pass cv=StratifiedKFold(n_splits=10, shuffle=True, random_state=0) to cross_val_score.

Show solution

Python

from sklearn.datasets import load_breast_cancer
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import StratifiedKFold, cross_val_score
from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import StandardScaler

X, y = load_breast_cancer(return_X_y=True)
cv = StratifiedKFold(n_splits=10, shuffle=True, random_state=0)
scores = cross_val_score(make_pipeline(StandardScaler(), LogisticRegression()), X, y, cv=cv)
print(round(scores.min(), 3), round(scores.max(), 3), round(scores.mean(), 3))

Output

0.947 1.0 0.977

6.Tune regularisation strength

Intermediate

Use GridSearchCV with 5 folds to choose C from [0.01, 0.1, 1, 10] for a scaled logistic regression on the breast cancer data. Print the best C and its cross-validated accuracy (3 decimals).

Show hint

In a pipeline made with make_pipeline, the parameter name is logisticregression__C.

Show solution

Python

from sklearn.datasets import load_breast_cancer
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import GridSearchCV
from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import StandardScaler

X, y = load_breast_cancer(return_X_y=True)
pipe = make_pipeline(StandardScaler(), LogisticRegression(max_iter=1000))
search = GridSearchCV(pipe, {"logisticregression__C": [0.01, 0.1, 1, 10]}, cv=5).fit(X, y)
print(search.best_params_["logisticregression__C"], round(search.best_score_, 3))

Output

1 0.981

7.Lower the threshold

Intermediate

Create imbalanced data with make_classification(n_samples=3000, weights=[0.95, 0.05], random_state=1) and split it 70/30 (stratified, random_state=0). Fit a logistic regression, then print the recall of class 1 at thresholds 0.5, 0.3 and 0.1 (3 decimals).

Show hint

Compare predict_proba(X_test)[:, 1] >= threshold with y_test using recall_score.

Show solution

Python

from sklearn.datasets import make_classification
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import recall_score
from sklearn.model_selection import train_test_split

X, y = make_classification(n_samples=3000, weights=[0.95, 0.05], random_state=1)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0, stratify=y)
proba = LogisticRegression(max_iter=1000).fit(X_train, y_train).predict_proba(X_test)[:, 1]
for threshold in [0.5, 0.3, 0.1]:
    print(threshold, round(recall_score(y_test, proba >= threshold), 3))

Output

0.5 0.633
0.3 0.796
0.1 0.918

Lowering the threshold flags more cases as class 1, so recall rises from 0.633 to 0.918. Precision falls in exchange; choose the threshold from the cost of each kind of mistake.

← All practice sets Learn Machine Learning with Python course →