Cluster Analysis

Jarccard distance: dist (data frame, method = "binary")

Categorical variables (e.g., true and false)

More than two categorical variables: 

• library(dummies)

• x <- dummy.data.frame(spring2023)

• dist(x, method= "binary")

Scale - standadization 

Types of Linkages in Clustering: complete, single, average. 

Hierarchical Clustering in R  (use the hierarchical clustering method to group observations)

• Visualizing the dendrogram (a tree diagram) 

• Cutting the tree 

• Application (more than 2 dimensions, continuous variables): market segmentation (i.e. use consumer characteristics to group them into subgroups).

k-means clustering

• What happens if you don't know the right value of the K in advance? - Visualize your plots. 

  • Elbow plot: total within-cluster sum of squares

  • Silhouette analysis method

    • Silhouette width consists of two parts: S(I)

      1. Within Cluster Distance: C(I)

      2. Closer Neighbour Distance: N(I)

    • S(I) values: 1 indicates well matches, 0 indicates on the border between two clusters, and -1 indicates a better fit within the neighboring cluster.

• Making sense of the k-means cluster