NEW
Font size
WorksheetsMCQs on Introduction to Clustering & Partitional Methods
Total questions: 60
Worksheet time: 30mins
What is the main goal of clustering?
Classification of data
Prediction of outcomes
Grouping similar data objects
Data compression
A cluster is defined as a collection of data objects that are
Randomly distributed
Similar within a cluster and dissimilar to other clusters
Identical in nature
Always numerical
Clustering is widely used in all EXCEPT
Image processing
Marketing research
Crime analysis
Compiler design
Which metric is fundamental for clustering algorithms?
Sorting order
Distance or similarity
Frequency count
Probability value
Clustering numeric data primarily uses
Logical operators
Geometric properties
Lexicographic ordering
Hash functions
Why is clustering categorical data difficult?
Too many values
Distance function is not well-defined
Requires normalization
Needs labeled data
Which of the following is NOT a stage of clustering process?
Data collection
Clustering tendency
Validation
Classification
Scalability is a requirement of clustering algorithms because
Data sets are very small
Data sets are very large
Algorithms are simple
Clusters are predefined
Which clustering paradigm forms a nested grouping of data?
Partitioning
Density-based
Hierarchical
Grid-based
The graphical representation of hierarchical clustering is called
Scatter plot
Histogram
Dendrogram
Decision tree
Agglomerative clustering follows which approach?
Top-down
Bottom-up
Randomized
Parallel
In agglomerative clustering, initially
One cluster exists
All objects are in the same cluster
Each object forms a separate cluster
Clusters are predefined
Divisive clustering follows which approach?
Bottom-up
Top-down
Greedy
Randomized
DIANA is an example of
Agglomerative clustering
Partitional clustering
Divisive clustering
Density-based clustering
A major disadvantage of hierarchical clustering is
High accuracy
Cannot handle numeric data
Once merged or split, it cannot be undone
Requires predefined k
Partitional clustering divides data into
Variable number of clusters
Predefined number of clusters
Overlapping clusters
Hierarchical clusters
The quality of partitional clustering is improved by
Sorting data
Swapping data points
Removing attributes
Increasing dimensions
Which of the following is a partitional clustering method?
DIANA
AGNES
K-Means
BIRCH
In partitional clustering, prototypes are
Always random
Never updated
Iteratively recomputed
Fixed throughout
The number of ways to partition n objects into k clusters is
n!
k!
kn/k!
nk
In K-Means, each cluster is represented by
Median
Mode
Mean
Boundary
The centroid in K-Means is
A random object
The mean of objects in the cluster
The farthest object
Always an actual data object
K-Means algorithm is sensitive to
Noise only
Duplicate data
Outliers
Missing values
Which step assigns objects to nearest centroid?
Initialization
Computation
Assignment
Validation
K-Means stops when
Maximum iterations reached
No change in cluster assignment
Error becomes zero
Data is sorted
K-Medoid represents each cluster using
Mean value
Median value
One actual object
Virtual point
PAM stands for
Partition and Merge
Partition Around Medoids
Parallel Analysis Model
Pattern Association Method
PAM works effectively for
Very large datasets
Streaming data
Small datasets
High-dimensional sparse data
PAM improves clustering quality by
Sampling
Swapping medoids
Merging clusters
Normalization
PAM has how many main modules?
One
Two
Three
Four
CLARA is designed to handle
Small datasets
Numerical data only
Larger datasets than PAM
Hierarchical data
CLARA applies PAM on
Entire dataset
Random clusters
Sample of the dataset
Sorted data
A weakness of CLARA is
High computation cost
Biased sampling
Inability to handle numeric data
Sensitivity to outliers
CLARANS is based on
Exhaustive search
Randomized search
Greedy search
Sequential search
CLARANS improves scalability by
Using full dataset
Avoiding medoids
Sampling neighbors dynamically
Fixed clustering
CLARANS stops when
All objects are clustered
Best global solution is found
A fixed number of local optima are found
Error becomes zero
Which algorithm is more scalable?
PAM
K-Means
CLARA
CLARANS
Which method uses mean as cluster center?
K-Medoid
PAM
K-Means
CLARA
Which algorithm cannot undo merge or split decisions?
K-Means
PAM
CLARA
Hierarchical clustering
Clustering tendency helps to
Label data
Decide number of clusters
Measure similarity
Store data
Clustering is a type of learning where
Supervised learning
Unsupervised learning
Semi-supervised learning
Reinforcement learning
Which of the following best describes clustering?
Assigning labels based on prior knowledge
Discovering hidden patterns without class labels
Predicting future values
Reducing dimensionality
In clustering, similarity between objects is generally measured using
Classification rules
Distance functions
Decision boundaries
Probability distributions
Which of the following is NOT a characteristic of a good clustering?
High intra-cluster similarity
Low inter-cluster similarity
Random grouping of objects
Compact clusters
Which data type is easiest to cluster using traditional distance measures?
Categorical
Text
Numerical
Boolean
One major challenge in clustering large databases is
Availability of labels
High dimensionality
Small dataset size
Lack of algorithms
Which factor strongly affects the outcome of clustering?
Data ordering
Choice of distance metric
File format
Storage device
Clustering differs from classification because clustering
Uses training data
Requires class labels
Does not require predefined classes
Is less accurate
The value of k in K-Means represents
Number of attributes
Number of iterations
Number of clusters
Number of data points
Choosing an inappropriate value of k may lead to
Perfect clustering
Overfitting or underfitting
Faster convergence
Reduced computation
Which distance measure is most commonly used in K-Means?
Manhattan distance
Euclidean distance
Cosine similarity
Jaccard coefficient
Initial selection of centroids in K-Means affects
Data storage
Execution speed only
Final clustering result
Input format
K-Means may converge to different results because
It is deterministic
It uses random initialization
It sorts data first
It uses categorical data
Which of the following is a limitation of K-Means?
Works only for hierarchical data
Cannot handle numerical data
Assumes spherical-shaped clusters
Requires labeled data
K-Means performs poorly when clusters
Are well separated
Have similar sizes
Are non-convex or irregular in shape
Are compact
Which situation makes K-Means unsuitable?
Numeric attributes
Presence of outliers
Low dimensional data
Small dataset size
The objective function minimized by K-Means is
Maximum distance between clusters
Sum of squared distances within clusters
Number of clusters
Number of iterations
In K-Means, reassignment of objects continues until
Maximum error is reached
No object changes its cluster
Data is normalized
All clusters have equal size
K-Means requires which type of input attributes?
Categorical only
Ordinal only
Numerical only
Mixed type
Which statement about K-Means is TRUE?
Cluster centers must be actual data points
Clusters can overlap
The algorithm guarantees global optimum
Results depend on initial centroids
