Resources

• Package kknn

Geographic clustering of UK cities

Animated example: https://towardsdatascience.com/the-5-clustering-algorithms-data-scientists-need-to-know-a36d136ef68

Exercise: Kmeans

Apply kmeans to to the iris dataset with 2, 3, and 4 clusters. Produce three scatter plots, with the points colored according to cluster assignment.

hdbscan

A fairly new alternative to kmeans, hdbscan does not require you to specify the number of categories to be assigned. It only requires a decision as to the minimum number of points needed to be included in a cluster. This minimum number acts as a smoothing parameter (such as a density bandwidth parameter or a histograms bin/bar width), with lower values finding more clusters. Other advantages of hdbscan include .

Exercise: Apply kmeans to the moons dataset and compare the results.

– Be sure to try different numbers of centers.

Exercise: Apply hdbscan to the moons dataset with different minimums for the number of points.

Exercise: Apply both kmeans and hdbscan to the ChickWeight dataset’s “weight” “Time” variables, and see how well you can get each to perform.

US Census Data

• US Census Data (Clustering) – Clustering based on demographics is a tried and true way to perform market research and segmentation.

Links

• Using clusterlab to benchmark clustering algorithms