Features:
1. Spatial validation
Spatial validation provides a collection of validation measures including (1) fragmentations (entropy, simpson), (2) join count ratio, (3) compactness (isoperimeter quotient) and (4) diameter.
Fragmentation is a measure of spatial validation of clusters. It includes:
- entropy, which measures the fraction of observations in each cluster
- entropy*, which is the standardized entropy measure
- simpson, which is an index for diversity measure in each cluster
- simpson*, which is the standardized simpson measure
For non-spatially constrained clusters, the validation also reports cluster_fragmentation, which is a list of Fragmentation objects for each cluster, or None for spatially constrained clusters.
JoinCountRatio is measure of join counts (the number of times a category is surrounded by neighbors of the same category) over the total number of neighbors after converting each category to a dummy variable. It includes:
- neighbors, the total number of neighbors of elements in a cluster
- join count, the total join count of elements in a cluster
- ratio: the ratio of total join count over total neighbors
Compactness is a measure of isoperimeter quotient for each spatially constrained cluster. It includes:
- area, the area of a cluster. For points, the convex hull is used to compute the area.
- perimeter, the perimeter of a cluster. For points, the convex hull is used to compute the perimeter
- isoperimeter_quotient, (4 * pi * area) / (perimeter^2)
Diameter is a measure of the longest shortest distance between any pairs in a cluster. It includes:
- steps, the longest shortest distance between any pairs
- ratio, the ratio of steps over the number of elements in the cluster
2. Make Spatial
Make spatially constrained clusters from spatially non-constrained clusters using the contiguity information from the input weights
3. SC K Means
Spatially constrained K Means clustering algorithm implemented.
4. Join Count Ratio for Unique Value Maps
Add join count ratio option for unique value maps
5. Block Weights
Add block weights option in the contiguity weights creation
6. Cluster Map Match
Compare two clustering results using cluster map match.
7. Windows Installer for lab environment
Create installer for Windows 7+, which can be used in a Windows Lab Environment.
8. Add installer for Apple M1
Fixes:
- Weights creation from polygons wrong: double precision epsilon was not handled in weights creation code.
- fix issue that 3D plot doesn't work on Mac OS with version > 1.18.0.6
- Update GeoDa-win7+.iss for issue MS 2015-2019 Redist problems #2360