CAP4770 Exam 2

T/F - Hierarchical clustering requires a predetermined number of clusters.

False

T/F - Normalization is essential in clustering to ensure all variables contribute equally to the distance measures.

True

T/F - K-means clustering assigns records to clusters based on probabilities.

False

T/F - Binary similarity measures are preferred when working with continuous data.

False

Which of the following is an unsupervised learning task?

• Clustering

• Regression

• Classification

• Reinforcement learning

Clustering

Which distance measure is most commonly used in clustering but has limitations such as sensitivity to outliers?

• Cosine similarity

• Manhattan distance

• Euclidean distance

• Jaccard similarity

Euclidean distance

In clustering, normalization of numerical variables ensures that:

• Larger variables dominate the clustering process

• All variables contribute equally to the distance measures

• Only important variables contribute to distance measures

• Clustering becomes faster

All variables contribute equally to the distance measures

In k-means clustering, the number of clusters is:

• Randomly chosen after clustering

• Automatically determined by the algorithm

• Predefined before running the algorithm

Predefined before running the algorithm

What does hierarchical clustering require that can make it computationally expensive?

• An n x n distance matrix

• A small dataset

• A random seed

• A predetermined number of clusters

An n x n distance matrix

T/F - The k-means algorithm always guarantees the globally optimal clustering solution.

False

T/F - K-means clustering is best suited for datasets with spherical clusters.

True

T/F - K-means is more sensitive to outliers compared to K-medoids clustering.

True

T/F - In centroid linkage clustering, the distance between clusters is calculated by averaging the distances between all points in the two clusters.

False

T/F - The sum of squared errors (SSE) decreases as the number of clusters (k) increases in k-means clustering.

True

T/F - The “elbow method” is used to determine the most appropriate number of clusters in k-means clustering.

True

T/F - K-means clustering can be applied to both continuous and categorical data without modification.

False

What is the main goal of k-means clustering?

• Minimize between-cluster variance

• Maximize between-cluster variance

• Minimize within-cluster variance

• Maximize within-cluster variance

Minimize within-cluster variance

Which of the following statements about K-medoids is true?

• It only uses squared Euclidean distance to calculate distances

• It is less sensitive to outliers compared to k-means

• It uses centroids to define clusters

• It assumes clusters are spherical

It is less sensitive to outliers compared to k-means

In k-means clustering, what is the purpose of the “elbow method”?

• To determine the optimal number of clusters

• To minimize outliers

• To increase the number of centroids

• To reduce the within-cluster variance

To determine the optimal number of clusters

Maximum Coordinate Distance is the same as _________. Both refer to the distance between two points where only the maximum absolute difference across any of the dimensions is considered.

• Euclidean distance

• Manhattan distance

• Chebyshev distance

• City block distance

Chebyshev distance

T/F - In agglomerative hierarchical clustering, the clustering process starts with each instance as an individual cluster.

True

T/F - Divisive clustering is a bottom-up approach to hierarchical clustering.

False

T/F - Single linkage clustering calculates the maximum distance between points in two clusters.

False

T/F - Ward’s method minimizes the loss of information at each step by using Error Sum of Squares (ESS).

True

T/F - Dendrograms are tree-like diagrams used to show the order of clustering and the distance between clusters.

True

In complete linkage clustering, how is the distance between two clusters measured?

• By the average distance between all pairs of points

• By the distance between the farthest points

• By the distance between centroids

• By the distance between the closest points

By the distance between the farthest points

What characteristic is commonly associated with single linkage clustering?

• Spherical clusters

• Large compact clusters

• Globular clusters

• Elongated, chain-like clusters

Elongated, chain-like clusters

Which of the following instances is most similar to A?

• B

• C

• D

• E

B

In the figure, two thresholds are represented by a solid red line and a dotted clue line. How many clusters would be formed at the solid red line threshold and at the dotted blue line threshold, respectively?

• 5, 5

• 5, 7

• 6, 11

• 7, 7

5, 7

Which of the following is NOT a feature of DBSCAN?

• Handles noise and outliers

• Discover clusters of arbitrary shapes

• Sensitive to initial cluster centroids

• Does not require you to pre-specify the number of clusters

Sensitive to initial cluster centroids

In DBSCAN, a point is classified as a core point if:

• It has at least a specified number of points within the epsilon radius

• It is on the boundary of the cluster

• It does not belong to any cluster

• It has a large distance to its nearest neighbors

It has at least a specified number of points within the epsilon

What is the main parameter that defines the neighborhood of a point in DBSCAN?

• Variance

• Eps (epsilon)

• k-nearest neighbords

• Number of clusters

Epsilon

Which of the following methods does DBSCAN use to form clusters?

• Identifying regions of high data density

• Maximizing intra-cluster similarity

• Minimizing distances to centroids

• Dividing data into equal partitions

Identifying regions of high data density

T/F - In DBSCAN, all border points have more than MinPts neighbors within the Eps radius.

False

T/F - DBSCAN can find clusters of arbitrary shapes, unlike k-means, which assumes spherical clusters.

True

T/F - DBSCAN requires the number of clusters to be specified before clustering begins.

False

T/F - The elbow method helps in determining the optimal number of clusters for k-means clustering.

True

T/F - External indices require ground truth labels to evaluate clustering results.

True

T/F - A higher Dunn Index indicates poor clustering quality.

False

In a GMM, the “soft clustering: property means that:

• Each data point is assigned a probability of belonging to each cluster

• Each data point is assigned to exactly one cluster

Each data point is assigned a probability of belonging to each cluster

T/F - the Silhouette Score ranges from 0 to 1, where 1 indicates the worst clustering quality.

False

What does the Dunn Index measure?

• Cohesion and dispersion

• Silhouette coefficient

• Intra-cluster compactness and inter-cluster separation

• Cluster variance

Intra-cluster compactness and inter-cluster separation

Which of the following is an internal index for cluster evaluation?

• Precision

• Purity

• Accuracy

• Silhouette Index

Silhouette Index

What does a high Silhouette Score indicate about clusters?

• Clusters overlap significantly

• Clusters are well-separated and cohesive

• Clusters are poorly defined

• Clusters are compact but not well-separated

Clusters are well-separated and cohesive

Which of the following best describes the purpose of a Gaussian Mixture Model?

• To perform hierarchical clustering by combining clusters in a tree-like cluster

• To separate data into clusters using a distance-based metric

• To model data with a single Gaussian distribution

• To model data with overlapping clusters by representing each cluster as a Gaussian distribution

To model data with overlapping clusters by representing each cluster as a Gaussian distribution