Data‑driven segmentation often starts with clustering algorithms. See if you can spot the key concepts behind modern cluster analysis.
The default distance measure for k‑means on standardised numeric data is ______ distance.
Cosine
Euclidean
Manhattan
Hamming
Euclidean distance minimises the sum of squared errors and aligns with the mean‑based centroid update.
In hierarchical clustering, the method that merges clusters based on the smallest average pairwise distance is called ______.
complete linkage
average linkage
single linkage
Ward’s method
Average linkage (UPGMA) balances cluster chaining and compactness by using mean inter‑cluster distance.
The elbow method helps analysts decide the optimal ______.
distance metric
number of clusters (k)
scaling factor
principal components
Plotting within‑cluster SSE against k reveals a point where marginal improvement drops—an ‘elbow’.
Standardising variables before clustering prevents ______ variables from dominating the distance calculation.
categorical
irrelevant
binary
high‑scale
Z‑scores equalise units so price in dollars does not swamp satisfaction in points.
K‑prototypes handles mixed data by combining numeric distance with ______ distance for categorical variables.
Hamming
Mahalanobis
Jaccard
Cosine
Hamming counts attribute mismatches, integrating seamlessly into the prototype update routine.
The silhouette coefficient jointly measures cluster cohesion and ______.
kurtosis
skewness
separation
sparsity
Values near 1 indicate points are closer to their own cluster than to neighbouring clusters.
In agglomerative clustering, merging continues until ______ cluster(s) remain.
one
a predefined silhouette score
k+1
zero
The dendrogram starts with n singletons and, if not cut earlier, finishes with a single mega‑cluster.
Probabilistic centre initialisation that improves k‑means consistency is known as ______.
fuzzy c‑means
random swap
Hartigan optimisation
k‑means++
k‑means++ spreads initial centroids, reducing the chance of poor local minima.
DBSCAN labels observations in sparse regions as ______.
linkages
centroids
medoids
noise
Points with insufficient neighbours within ε are treated as outliers rather than forced into clusters.
Ward’s method in hierarchical clustering aims to minimise the total ______ within clusters at each step.
dimensionality
entropy
distance metric
variance
By joining clusters that produce the smallest increase in SSE, Ward’s creates compact, equal‑size groups.
Starter
Keep exploring the fundamentals.
Solid
Great job—refine the finer points and you’ll lead the pack.
Expert!
You’ve mastered this segmentation concept—apply it boldly.
