Hierarchical Geo Clustering

Install

The library uses the cgal library, if the library is use inside a conda enviroment the installation procedure should create the enviroment and install the cgal library using conda install -c conda-forge cgal.

To install the library can be simply:

pip install git+https://github.com/CentroGeo/HierarchicalGeoClustering.git

How to use

Build a TreeCluster

We can build a hierarchical geographical cluster, the TreeCluster class allows to create te structure and then populate it.

HGC = TreeClusters(3, random_seed= 12)

HGC.populate_tree(number_per_cluster=100)

HGC.print_structure()

Root
Root    
├── Root_l_0_n_0
│   ├── Root_l_0_n_0_l_1_n_0
│   │   ├── Root_l_0_n_0_l_1_n_0_l_2_n_0
│   │   ├── Root_l_0_n_0_l_1_n_0_l_2_n_1
│   │   ├── Root_l_0_n_0_l_1_n_0_l_2_n_2
│   │   ├── Root_l_0_n_0_l_1_n_0_l_2_n_3
│   │   └── Root_l_0_n_0_l_1_n_0_l_2_n_4
│   └── Root_l_0_n_0_l_1_n_1
│       ├── Root_l_0_n_0_l_1_n_1_l_2_n_0
│       └── Root_l_0_n_0_l_1_n_1_l_2_n_1
├── Root_l_0_n_1
│   ├── Root_l_0_n_1_l_1_n_0
│   │   └── Root_l_0_n_1_l_1_n_0_l_2_n_0
│   ├── Root_l_0_n_1_l_1_n_1
│   │   ├── Root_l_0_n_1_l_1_n_1_l_2_n_0
│   │   ├── Root_l_0_n_1_l_1_n_1_l_2_n_1
│   │   └── Root_l_0_n_1_l_1_n_1_l_2_n_2
│   ├── Root_l_0_n_1_l_1_n_2
│   ├── Root_l_0_n_1_l_1_n_3
│   │   ├── Root_l_0_n_1_l_1_n_3_l_2_n_0
│   │   └── Root_l_0_n_1_l_1_n_3_l_2_n_1
│   └── Root_l_0_n_1_l_1_n_4
│       ├── Root_l_0_n_1_l_1_n_4_l_2_n_0
│       ├── Root_l_0_n_1_l_1_n_4_l_2_n_1
│       └── Root_l_0_n_1_l_1_n_4_l_2_n_2
└── Root_l_0_n_2
    ├── Root_l_0_n_2_l_1_n_0
    │   ├── Root_l_0_n_2_l_1_n_0_l_2_n_0
    │   └── Root_l_0_n_2_l_1_n_0_l_2_n_1
    ├── Root_l_0_n_2_l_1_n_1
    │   ├── Root_l_0_n_2_l_1_n_1_l_2_n_0
    │   ├── Root_l_0_n_2_l_1_n_1_l_2_n_1
    │   ├── Root_l_0_n_2_l_1_n_1_l_2_n_2
    │   ├── Root_l_0_n_2_l_1_n_1_l_2_n_3
    │   └── Root_l_0_n_2_l_1_n_1_l_2_n_4
    ├── Root_l_0_n_2_l_1_n_2
    │   ├── Root_l_0_n_2_l_1_n_2_l_2_n_0
    │   ├── Root_l_0_n_2_l_1_n_2_l_2_n_1
    │   └── Root_l_0_n_2_l_1_n_2_l_2_n_2
    └── Root_l_0_n_2_l_1_n_3
        ├── Root_l_0_n_2_l_1_n_3_l_2_n_0
        ├── Root_l_0_n_2_l_1_n_3_l_2_n_1
        ├── Root_l_0_n_2_l_1_n_3_l_2_n_2
        ├── Root_l_0_n_2_l_1_n_3_l_2_n_3
        └── Root_l_0_n_2_l_1_n_3_l_2_n_4

fig, axs = plt.subplots( figsize=(8,8))
HGC.visualize(axs, polygon=True)

HGC.root.polygon_cluster

HGC.get_deepth()

4

Extract the points and clusterize

To clusterize we extract the points and use the clusterize tool implemented in the Clustering module

original_points= HGC.get_points_tree()
X_2=np.array([[p.x,p.y] for p in original_points])
dic_points={'points':[X_2], 'parent':''}

HGC_adapta_DBSCAN = recursive_clustering_tree(dic_points,
                                       levels_clustering = 3,
                                       algorithm = 'adaptative_DBSCAN'
                                      )

fig, axs = plt.subplots( figsize=(8,8))
HGC_adapta_DBSCAN.visualize(axs, polygon=True)

Get the Similarity Shape Measurement

To obtain the SSM the clusters tree has to be used, this due to the use of the geometric shape in the metric. Therefore the polygons of each cluster on each level has to be able to compare.

form_metric=[]

for l in range(0, 4):
    d = { 'Level': l,                              
        'adaptive DBSCAN':SSM(HGC.levels_nodes[l],
                            HGC_adapta_DBSCAN.levels_nodes[l])
        }
    form_metric.append(d)

form_metric

[{'Level': 0, 'adaptive DBSCAN': 0.6046098496677557},
 {'Level': 1, 'adaptive DBSCAN': 0.6195370749041116},
 {'Level': 2, 'adaptive DBSCAN': 0.4080125145736268},
 {'Level': 3, 'adaptive DBSCAN': 0.22732397098600046}]

df_metric_form = pd.DataFrame(form_metric)

df_metric_form

<style scoped> .dataframe tbody tr th:only-of-type { vertical-align: middle; }

.dataframe tbody tr th {
    vertical-align: top;
}

.dataframe thead th {
    text-align: right;
}

</style>

	Level	adaptive DBSCAN
0	0	0.604610
1	1	0.619537
2	2	0.408013
3	3	0.227324

A detail documentation is available