Explore Map Partition Using CLustering with Interaction Using R Shiny

Background

The number of US congressional districts are allocated on a state by state basis according to the total number of house seats fixed at 435 and the result of the US census conducted every ten years.

The purpose is to consider how to do absolutely fair congressional district apportionment using clustering and plotting capability of R. It is the antithesis of gerrymander.

Approach

Consider K-means clustering and heirarchical clustering. This is an exploration of it. The application uses heirarchical clustering.

The user interace provides option for user to enter:

• A state
• Clustering method parameter
  
• Clusterlink parameter
  used in the hclust routine.

The result is drawn on a map and on a dendro plot.

These libraries are used:

• 'xlsx','plyr','amap','tmap','maptools'

Data Sets

• Apportionment by State US Census, how many districts per state
• Population data US USDA for map data and population data to be used in future version doing wieghted clustering operation.
• Geo Tiger Files US Census Geo shape files of the USA

Data Preparation

• Data sets merge into the shape files
• After the shape file data set is built, an extract is done to run the clustering
• A masster Shape Data Frame is built with from the original tiger file, apportionment data and population data
• FIPS links Shape to population
• State name links shape and apportionment
• At runtime the master data set if filtered on state name.

Setup

• THe form has two frames, sidebar frame and main frame.
• The sidebar frame has drop downs for
  • State
  • Cluster Method Param
  • Cluster Link Param
• The main Frame
  • Report the selected option
  • Plots Colorpleth State map of counties
  • Plots Heirarchical dendro plot with threshold and boxing.