Cluster analysis is an important tool for unsupervised learning. Determining the number of clusters K is a difficult problem. Our goal is to explore the different ways which are currently used to determine the optimal number of clusters in k-means.

###Team members:

• Nikhila Balaji
• Katherine Brey
• Hussein Koprly
• Alec McDivitt
• Julie Persinger
• Hunter Sipe
• Yi Wei

####Mentor: Bharathkumar “Tiny” Ramachandra

###Objective:

• Learn how to think about the solution to a hard problem in unsupervised learning, implement the solution, create and execute a structured work plan
• Identify/Recognize some challenges that arise in unsupervised learning and understand why determining the optimal number of clusters is a hard problem
• Analyze and identify the steps in the code structure in the k-means clustering implementation
• Be able to generate 4 clustering datasets with different properties
• Model the mathematical representation of K in all the 4 methods:
  1. Gap Statistic
  2. Elbow
  3. Information theoretic
  4. Avg Silhouette

###Tasks - Implementations:

• Implementing K-means
• Gap Statistic
• Elbow method
• Information Theoretic
• Avg Silhouette

####Required Libraries:

install.packages('plot3D')
install.packages('scatterplot3d')
install.packages('car')
install.packages('pracma')