Exploratory Data Analysis

• hypothesis creation
• data patterns revelation
• exploration, not visualization
• statistics

ESDA tools

scatter plot

ESDA tools

parallel coordinate plot

ESDA tools

brushing

• connected display of spatial and tabular/statistical data

 

paneling

• one visualization for multiple subsets of data

ESDA tools

k-means clustering

• given: data 
• points are nearer to the center in "their" cluster
• each cluster has an unknown "center" - how do you choose one?

ESDA tools

k-means clustering

• sum distances to center over clusters - try to minimize this value
• only numeric attributes, yields different results

1. initialize centers (from 1 to k)
2. for each point:
   1. find nearest centroid
   2. assign the point to cluster
3. for each centroid:
   1. calculate new position based on average of attribute values
4. stop when no points change memberships

ESDA tools

k-means clustering

• Cluster 3.0 
• http://bonsai.hgc.jp/~mdehoon/software/cluster/software.htm

 

can you think of any disadvantages of this algorithm?

ESDA tools

principal component analysis (PCA)

• directions where there is the most variance
• eigenvectors = directions
• eigenvalue = how much variance there is in eigenvector
• principal component = eigenvector with the highest eigenvalue
• eigenvectors put the data into a new set of dimensions
• can be used to reduce the dimensions of a data set

ESDA tools

• ESTAT
• CCmaps
• GeoDa
• GeoVIZ
• CommonGIS
• TimeMap
• https://walkerke.shinyapps.io/neighborhood_diversity/