SciPy with scikit-learn pipeline - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of a simple pipeline with StandardScaler and LogisticRegression

What is the output of the following code snippet that creates a pipeline with a scaler and logistic regression, then fits and predicts on a test sample?

SciPy

from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression
import numpy as np

X_train = np.array([[1, 2], [2, 3], [3, 4], [4, 5]])
y_train = np.array([0, 0, 1, 1])

pipeline = Pipeline([
    ('scaler', StandardScaler()),
    ('logreg', LogisticRegression(random_state=0))
])

pipeline.fit(X_train, y_train)

X_test = np.array([[1.5, 2.5]])
prediction = pipeline.predict(X_test)
print(prediction)

A [1]

B [1 0]

C [0 1]

D [0]

💡 Hint

Think about how StandardScaler transforms the input and how logistic regression predicts based on training labels.

The pipeline first scales the input features. The logistic regression model trained on the scaled training data predicts class 0 for the test input [1.5, 2.5].

❓ data_output

intermediate

Shape of transformed data after applying PCA in a pipeline

Given the following pipeline that applies PCA to reduce dimensionality, what is the shape of the transformed data after calling transform on X_test?

SciPy

from sklearn.pipeline import Pipeline
from sklearn.decomposition import PCA
import numpy as np

X_train = np.random.rand(10, 5)
X_test = np.random.rand(3, 5)

pipeline = Pipeline([
    ('pca', PCA(n_components=2))
])

pipeline.fit(X_train)
X_transformed = pipeline.transform(X_test)
print(X_transformed.shape)

A (3, 5)

B (3, 2)

C (10, 2)

D (10, 5)

💡 Hint

PCA reduces the number of features to n_components but keeps the number of samples the same.

The transform method returns data with the same number of samples (3) but reduced features (2). So the shape is (3, 2).

🔧 Debug

advanced

Identify the error in pipeline usage with SciPy function

What error will this code raise when trying to use a SciPy function inside a scikit-learn pipeline step?

SciPy

from sklearn.pipeline import Pipeline
from sklearn.preprocessing import FunctionTransformer
from scipy.special import expit
import numpy as np

X = np.array([[0, 1], [2, 3]])

pipeline = Pipeline([
    ('sigmoid', FunctionTransformer(expit)),
])

pipeline.fit(X)
output = pipeline.transform(X)
print(output)

A No error, outputs transformed array

B AttributeError: 'FunctionTransformer' object has no attribute 'fit'

C TypeError: 'FunctionTransformer' object is not callable

D ValueError: Input contains NaN

💡 Hint

FunctionTransformer wraps a function to apply it during transform. Check if expit works element-wise.

FunctionTransformer correctly applies the SciPy expit function element-wise. No error occurs and the output is the sigmoid transformed array.

🚀 Application

advanced

Using a custom SciPy statistical function in a pipeline step

You want to add a pipeline step that replaces each feature with its z-score using SciPy's zscore function. Which pipeline step code correctly applies this transformation?

A ('zscore', StandardScaler())

B ('zscore', FunctionTransformer(scipy.stats.zscore))

C ('zscore', FunctionTransformer(lambda x: scipy.stats.zscore(x, axis=0)))

D ('zscore', FunctionTransformer(lambda x: scipy.stats.zscore(x, axis=1)))

💡 Hint

zscore needs axis=0 to standardize features column-wise.

Option C correctly wraps scipy.stats.zscore with axis=0 to standardize features. Option C misses axis argument, default is axis=0 but may cause issues. Option C uses StandardScaler which is similar but not SciPy. Option C uses axis=1 which standardizes rows, not features.

🧠 Conceptual

expert

Why integrate SciPy functions in scikit-learn pipelines?

What is the main advantage of integrating SciPy functions inside scikit-learn pipelines using FunctionTransformer?

A It allows seamless combination of SciPy transformations with scikit-learn estimators for consistent preprocessing and model fitting.

B It automatically converts SciPy functions into scikit-learn estimators with fit and predict methods.

C It speeds up SciPy functions by compiling them into C code within the pipeline.

D It enables SciPy functions to handle missing data automatically during pipeline execution.

💡 Hint

Think about how pipelines help organize multiple steps in machine learning workflows.

FunctionTransformer wraps SciPy functions so they can be used as pipeline steps, allowing smooth chaining of preprocessing and modeling steps in scikit-learn.