Why clustering groups similar data in SciPy - Challenge Your Understanding

Challenge - 5 Problems

🧠 Conceptual

intermediate

Why does clustering group similar data points?

Which of the following best explains why clustering algorithms group similar data points together?

A Clustering algorithms group data points based on their proximity in feature space, so points close to each other are grouped together.

B Clustering algorithms randomly assign data points to groups without considering their features.

C Clustering algorithms group data points by sorting them alphabetically based on their labels.

D Clustering algorithms group data points by their order of appearance in the dataset.

💡 Hint

Think about how distance or similarity between points affects grouping.

Clustering groups data points that are close to each other in the feature space because they share similar characteristics. This proximity is measured using distance metrics like Euclidean distance.

❓ Predict Output

intermediate

Output of clustering labels using scipy

What is the output labels array after running this clustering code?

SciPy

from scipy.cluster.hierarchy import fcluster, linkage
import numpy as np

# Sample data points
X = np.array([[1, 2], [1, 4], [1, 0], [10, 2], [10, 4], [10, 0]])

# Perform hierarchical clustering
Z = linkage(X, method='single')

# Form flat clusters with max distance 3
labels = fcluster(Z, t=3, criterion='distance')
print(labels)

A [1 1 2 2 3 3]

B [1 2 3 4 5 6]

C [2 2 2 1 1 1]

D [1 1 1 2 2 2]

💡 Hint

Look at how points close in space are grouped with a distance threshold of 3.

The first three points are close together and form cluster 1. The last three points are close and form cluster 2. The labels array reflects these two groups.

❓ data_output

advanced

Number of clusters formed with different distance thresholds

Given the same data and linkage matrix, how many clusters are formed when the distance threshold changes?

SciPy

from scipy.cluster.hierarchy import fcluster, linkage
import numpy as np

X = np.array([[1, 2], [1, 4], [1, 0], [10, 2], [10, 4], [10, 0]])
Z = linkage(X, method='single')

clusters_t2 = fcluster(Z, t=2, criterion='distance')
clusters_t5 = fcluster(Z, t=5, criterion='distance')

num_clusters_t2 = len(set(clusters_t2))
num_clusters_t5 = len(set(clusters_t5))
print(num_clusters_t2, num_clusters_t5)

A 2 2

B 3 2

C 3 1

D 1 3

💡 Hint

Smaller distance thresholds create more clusters; larger thresholds merge clusters.

At threshold 2, three clusters form because points are grouped tightly. At threshold 5, two clusters form as some groups merge.

❓ visualization

advanced

Interpreting a dendrogram for clustering

Which statement correctly describes the dendrogram shown below for hierarchical clustering?

(Imagine a dendrogram with two main branches splitting at a height around 3)

A The dendrogram shows two main clusters formed when cutting at height 3, grouping similar points together.

B The dendrogram shows clusters formed by sorting points alphabetically.

C The dendrogram indicates that no clusters can be formed because all points are too far apart.

D The dendrogram shows that all points are identical and form one cluster at any height.

💡 Hint

Look at where the branches join and the height to decide cluster groups.

A dendrogram visually represents how clusters merge at different distances. Cutting at height 3 splits data into two main clusters.

🔧 Debug

expert

Identify the error in clustering code

What error will this code raise when run?

SciPy

from scipy.cluster.hierarchy import linkage, fcluster
import numpy as np

X = np.array([[1, 2], [1, 4], [1, 0], [10, 2], [10, 4], [10, 0]])

Z = linkage(X, method='single')

# Incorrect use of fcluster with invalid criterion
labels = fcluster(Z, t=3, criterion='invalid')
print(labels)

A No error, prints cluster labels

B IndexError: list index out of range

C ValueError: criterion must be one of ['inconsistent', 'distance', 'maxclust']

D TypeError: linkage() missing required positional argument

💡 Hint

Check the valid options for the 'criterion' parameter in fcluster.

The 'criterion' parameter must be one of 'inconsistent', 'distance', or 'maxclust'. Using 'invalid' raises a ValueError.