Evaluation#

Below are two measurements for comparing clusterings. Both are commonly used together since their purpose is the same, but they have different theoretical bases. The value is 0 when the compared partitions are random and independent, and it is 1 when they are identical.

ARI#

To compare two attribute map clusterings, you can use the ARI (Adjusted Rand Index) function, which returns a real number between 0 and 1 indicating the degree of similarity between the two solutions.

Parámetros#

  • clustering_1: (Numpy Array) Contains the clusters to which each point belongs for clustering 1.

  • clustering_2: (Numpy Array) Contains the clusters to which each point belongs for clustering 2.

Return#

  • ARI: (float) ARI score.

   from SpatialCluster.metrics.ARI import ARI
   ari = ARI(clustering_1, clustering_2)

If you want to compare more than one pair of maps at once, you can generate a matrix that shows the ARI relationship for each pair of clusterings.

Parameters#

  • clusterings: (Pandas DataFrame) Contains the clusters to which each point belongs for the different clusterings to be compared.

  • plot: (bool) Indicates whether to display the matrix graphically.

Return#

  • ARI_matr: (Pandas DataFrame) Contains the ARI score for each pair of clusterings.

   from SpatialCluster.metrics.ARI import ARI_matrix
   ari_matr = ARI_matrix(clusterings, plot=True)

AMI#

To compare two attribute map clusterings, you can use the AMI (Adjusted Mutual Information) function, which returns a real number between 0 and 1 indicating the degree of similarity between the two solutions.

Parameters#

  • clustering_1: (Numpy Array) Contains the clusters to which each point belongs for clustering 1.

  • clustering_2: (Numpy Array) Contains the clusters to which each point belongs for clustering 2.

Return#

  • AMI: (float) AMI score.

   from SpatialCluster.metrics.AMI import AMI
   ami = AMI(clusters_map1, clusters_map2)

If you want to compare more than one pair of maps at once, you can generate a matrix that shows the AMI relationship for each pair of clusterings.

Parameters#

  • clusterings: (Pandas DataFrame) Contains the clusters to which each point belongs for the different clusterings to be compared.

  • plot: (bool) Indicates whether to display the matrix graphically.

Return#

  • AMI_matr: (Pandas DataFrame) Contains the AMI score for each pair of clusterings.

   from SpatialCluster.metrics.AMI import AMI_matrix
   ami_matr = AMI_matrix(clusterings, plot=True)