K-Means Clustering QUIZ (MCQ QUESTIONS AND ANSWERS)

The algorithm will fail to converge

The resulting clusters will be too broad and may not capture the underlying structure of the data

The resulting clusters will be too specific and may overfit the data

The algorithm will automatically adjust the number of clusters

Random initialization

K-means++

Running the algorithm multiple times with different initializations and selecting the best solution

Both B and C

To measure the compactness of clusters

To measure the separation between clusters

To evaluate the quality of the clustering solution

To determine the optimal number of clusters

By using the Gower distance

By one-hot encoding the categorical variables

By standardizing the continuous variables

All of the above

DBSCAN

Hierarchical clustering

K-modes

Spectral clustering

K-means assumes that all clusters have similar sizes

K-means is sensitive to outliers

K-means assumes that all clusters have a spherical shape

K-means is sensitive to the initial placement of cluster centroids

By identifying data points that are far from their cluster centroids

By identifying clusters with few data points

By identifying data points that have a high silhouette score

By identifying data points with a high between-cluster variance

The number of clusters

The distance metric used

The initialization method

All of the above

By performing principal component analysis (PCA) on the data

By using the cluster centroids as a reduced representation of the data

By applying t-distributed stochastic neighbor embedding (t-SNE)

By using the cluster assignments as new features

Randomly from the data points

By using the K-means++ initialization method

By employing a separate clustering algorithm

Both A and B

Use the Gower distance

One-hot encode the categorical variables

Replace K-means with K-medoids

All of the above

Yes, but it may require additional preprocessing

Yes, but only if an appropriate distance metric is used

No, K-means clustering assumes convex clusters

No, K-means clustering assumes linearly separable clusters

The point at which the centroids stop changing significantly

The point at which the algorithm has found the optimal number of clusters

The point at which the within-cluster sum of squares reaches a minimum

The point at which the algorithm has found the best distance metric

K-means is a partitional clustering method, while hierarchical clustering is a tree-based method

K-means is sensitive to outliers, while hierarchical clustering is robust to outliers

K-means requires the number of clusters to be specified, while hierarchical clustering does not

All of the above

The algorithm will fail to converge

The resulting clusters will be too specific and may overfit the data

The resulting clusters will be too broad and may not capture the underlying structure of the data

The algorithm will automatically adjust the number of clusters

Image segmentation

Text classification

Anomaly detection

Time series forecasting

To determine the optimal number of clusters

To identify the best distance metric

To select the best initialization method

To determine the convergence criteria

O(n)

O(n log n)

O(nkI)

O(n^2)

Continuous data

Categorical data

Binary data

Text data

It assumes clusters have a spherical shape

It cannot handle categorical data

It is sensitive to the initial placement of cluster centroids

All of the above

It reassigns the data points to the nearest non-empty cluster

It selects a new centroid for the empty cluster from the remaining data points

It removes the empty cluster from the final solution

It reinitializes the empty cluster centroid

Easy to implement and understand

Scalable to large datasets

Guaranteed to find the global optimum

Converges relatively quickly

By employing a more efficient clustering algorithm

By using an initialization technique that reduces the sensitivity to the initial placement of cluster centroids

By automatically determining the optimal number of clusters

By handling missing data

It measures the similarity between clusters

It measures the dissimilarity between clusters

It measures the compactness of clusters

It measures the separation between clusters

K-means uses centroids, while K-medoids use medoids

K-means is a hierarchical clustering algorithm, while K-medoids is a partitional clustering algorithm

K-means is sensitive to outliers, while K-medoids is robust to outliers

K-means can handle categorical data, while K-medoids cannot

Sensitivity to the initial placement of cluster centroids

Inability to handle missing data

Inability to handle categorical data

All of the above

Clusters have a spherical shape

Clusters have similar densities

Clusters have similar sizes

Clusters are linearly separable

By selecting the number of clusters that minimizes the within-cluster variance

By using domain knowledge or expert judgment

By employing an elbow plot or silhouette analysis

By using cross-validation

Dimensionality reduction

Classification

Regression

Partitioning data into clusters

Euclidean distance

Manhattan distance

Cosine similarity

Jaccard similarity