Machine Learning - Clustering

Here are some things that can be done with clustering:

• Make predictions in numerical or categorical form

• Finding patterns in data based on similarity

• Exploratory Data Analysis

• Outlier Detection

The basic differences between Clustering and Classification are:

The calculation of the proximity between the centroid and cluster members can be measured using the following distance calculation methods:

• Manhattan Distance

• Euclidean Distance

• Chebyshev Distance

• Mahalanobis Distance

Dbscan requires two parameters to run its model, namely:

• K for n of cluster

• R for radius of neighborhood

• M for minimum number of neighbor

• K for n of neighbor

Clustering is part of the model of Machine Learning which can be grouped into Supervised Learning and Unsupervised Learning.

• True

• False

DBScan clustering is good for Spherical clusters.

• True

• False

The use of clustering can be combined with classification, one of which is used for automated labeling of data.

• True

• False

Clustering performance with K-Means can be measured from the following:

• The average proximity of the centroid to the cluster members

• The more clusters, the better the results

• The fewer iterations of the modeling, the better the results

• Clustering cannot be measured

K-Means clustering is good for Aribitrary-Shape Clusters.

• True

• False

K-means is one of the clustering methods that can determine the number of cluster values or k values in the modeling process. To get the optimal cluster value, the following methods can be used:

• Centroid Method

• Elbow Method

• K-Means Method

• Random Method