04/03/2022

Objectives and Background

  • Determine if states can be grouped together
  • Clustering aid in future legislative actions
  • Show exploratory analysis and model fitting
  • Show the final model and provide recommendations
  • Data is of U.S. crime rates from 1973
  • Attributes: rape, murder, assault, and urban population

Correlation Plot

Histograms and Boxplots of Attributes

Sillhouette and Gap Statistics

Cluster Plot k=2 through k=5

Cluster Maps for k=2 through 4

The Final Models

Conclusions and Recommendations

  • Centroid based clustering analysis of U.S. crime rates and urban population data showed k=2 and k=4 are most stable
  • Cluster plots, Gap statistics, and silhouettes support models
  • Outliers were retained since the dataset was small
  • Clusters are separated by geographic regions
  • Recommend revisiting model every 5 years to prove the validity of the model
  • Recommend investigating other demographics of the state clusters (i.e. elections, poverty rates, or education rates)