Module 5: Clustering

2. Graded Quiz: Clustering

Question 1 The objective of k-means clustering is:

• Yield the highest out of sample accuracy

• Separate dissimilar samples and group similar ones

• Minimize the cost function via gradient descent

• Maximize the number of correctly classified data points

Question 2 Which option correctly orders the steps of k-means clustering?

1. Re-cluster the data points

2. Choose k random observations to calculate each cluster’s mean

3. Update centroid to take cluster mean

4. Repeat until centroids are constant

5. Calculate data point distance to centroids

• 2, 5, 3, 1, 4

• 2, 1, 4, 5, 3

• 3, 5, 1, 4, 2

• 2, 3, 4, 5, 1

Question 3 How can we gauge the performance of a k-means clustering model when ground truth is not available?

• • • Calculate the R-squared value to measure model fit.

    • Take the average of the distance between data points and their cluster centroids.

    • Calculate the number of incorrectly classified observations in the training set.

    • Determine the prediction accuracy on the test set.

Question 4 When the parameter K for k-means clustering increases, what happens to the error?

• • • It will decrease because the data points are less possible to be in the wrong cluster.

    • It will increase because incorrectly classified points are further from the correct centroid.

    • It might increase or decrease depending on if data points are closer to the centroid.

    • It will decrease because distance between data points and centroid will decrease.

Question 5 Which of the following is true for partition-based clustering but not hierarchical nor density-based clustering algorithms?

• • • Partition-based clustering produces sphere-like clusters.

    • Partition-based clustering can handle spatial clusters and noisy data.

    • Partition-based clustering is a type of unsupervised learning algorithm.

    • Partition-based clustering produces arbitrary shaped clusters.