Virtual Labs

Clustering: K-Means Algorithm

Why does adding more clusters (increasing K) always reduce WCSS?

a: Because clusters become larger Explanation

Explanation

b: Because each point gets closer to its assigned centroid Explanation

Explanation

c: Because distances are normalized Explanation

Explanation

d: Because centroids stop updating Explanation

Explanation

If a new feature that is highly correlated with an existing feature is added to a K-Means clustering dataset, what is the most likely outcome?

a: Clustering performance will always improve Explanation

Explanation

b: Cluster centroids will become identical Explanation

Explanation

c: The new feature may introduce redundancy without improving cluster quality Explanation

Explanation

d: WCSS will automatically become zero Explanation

Explanation

Why does K-Means assume clusters are convex (spherical)?

a: Because it uses linear regression Explanation

Explanation

b: Because distance is measured from a single centroid Explanation

Explanation

c: Because it uses probability distributions Explanation

Explanation

d: Because it sorts data Explanation

Explanation

What happens if K is chosen too small?

a: Clusters become too specific Explanation

Explanation

b: Different groups may be merged into one cluster Explanation

Explanation

c: WCSS becomes zero Explanation

Explanation

d: Algorithm diverges Explanation

Explanation

Why does K-Means require multiple runs with different initializations?

a: To reduce dataset size Explanation

Explanation

b: To avoid poor local minima and improve solution quality Explanation

Explanation

c: To change number of features Explanation

Explanation

d: To normalize data Explanation

Explanation

Why can K-Means fail when clusters have different densities and sizes simultaneously?

a: Because it assumes uniform density and size Explanation

Explanation

b: Because it ignores centroids Explanation

Explanation

c: Because it cannot compute distances Explanation

Explanation

d: Because it removes small clusters Explanation

Explanation

What does a Silhouette score close to 1 signify about the clustering result?

a: Data points are incorrectly assigned Explanation

Explanation

b: Clusters are dense and well separated Explanation

Explanation

c: Clusters are highly overlapping Explanation

Explanation

d: All clusters have equal number of samples Explanation

Explanation

What is the main idea behind the Elbow Method in clustering?

a: To find minimum WCSS Explanation

Explanation

b: To identify the point where adding more clusters gives little improvement Explanation

Explanation

c: To maximize silhouette score Explanation

Explanation

d: To reduce dataset size Explanation

Explanation

What does a lower Davies–Bouldin score indicate?

a: Poor clustering with high overlap Explanation

Explanation

b: Better clustering with compact and distinct clusters Explanation

Explanation

c: High variance within clusters Explanation

Explanation

d: Larger number of clusters Explanation

Explanation

What does a low Silhouette score indicate?

a: Strong cluster separation Explanation

Explanation

b: Points are near decision boundaries or misclassified Explanation

Explanation

c: Perfect clustering Explanation

Explanation

d: Equal cluster sizes Explanation

Explanation