Data Analytics Class Test 5

Identify an unsupervised machine learning algorithm

• Logistic Regression

• K Means Clustering

• Linear Regression

• Support Vector Machine

K means algorithm is based on ________selection.

• hyperplane

• seed point

• centroid

• support vector

Say True or False: The objective of k-means algorithm is to maximize the sum of distances between the data point and their corresponding clusters.

• True

• False

The K in the K-Means algorithm specifies which of the following:

• The number of data-points that we want to cluster out of a larger set of data-points.

• The average distance between cluster centroids over all algorithm iterations

• The number of partitions(clusters) that we want to get out of a given data-set

• This is the number of data-points that the similarity metric considers at each iteration.

Every iteration of the K-Means algorithm contains which of the following steps:

• Randomly assigning all data-points to one of K clusters

• Assigning data-points to the closest centroid using a given similarity (distance) measure.

• Randomly assigning the positions of K centroids in the data-point space.

• Calculate the average Euclidean distance between all cluster centroids.

The distance (similarity) function used by K-Means does which of the following:

• Computes the average distance between all of n real-valued data-points in a given data-set

• Converts a given a set of n real-valued data-points into a vector: x1,. . ., xn, of integer values

• Implements a distance calculation, dist(xi,mj), between each data-point xi and each cluster centroid (mj)

• Calculates the average Euclidean distance between K cluster centroids in the n dimensional space of all data

The convergence criterion of the K-Means algorithm is usually (in the simplest implementation of K-Means), which of the following:

• When the Sum of Squared Errors(SSE) stops decreasing

• When the algorithm outputs K clusters at any given iteration

• After n (user specified) iterations

• When the cluster centroids are no longer changing position and all data-points have been assigned to one of the K clusters (i.e. no more data-point re-assignments)

The main difference between clustering and classification is which of the following:

• Classification algorithms only use artificial neural networks.

• Clustering algorithms are supervised machine-learning algorithms whereas classification algorithms are unsupervised machine-learning algorithms.

• Clustering algorithms use an error function known as a similarity (distance) metric,whereas there are no error functions in classification algorithms.

• Clustering uses unlabelled training data to extract patterns (clusters) from any given data-set, whereas classification uses labelled training data to distinguish between (and thus classify) categories in the data.

Say True of False: In k means algorithm, intracluster distance should be large.

• True

• False

What is the other name of centroid in K Means Clustering Algorithm?