Network classification

Classify road network into one of following labels: - town - suburb - village

[1]:

from blocksnet.enums import SettlementCategory
import os
import osmnx as ox
from tqdm import tqdm

graphs = []

for pp in SettlementCategory:
    folder_path = pp.value
    for filename in tqdm(os.listdir(folder_path), desc=pp.value):
        file_path = os.path.join(folder_path, filename)
        g_nx = ox.load_graphml(file_path)
        g_nx.graph['category'] = pp.value
        graphs.append(g_nx)

town: 100%|██████████| 184/184 [00:15<00:00, 12.21it/s]
suburb: 100%|██████████| 122/122 [00:03<00:00, 38.13it/s]
village: 100%|██████████| 49/49 [00:00<00:00, 54.34it/s]

[2]:

from blocksnet.machine_learning.strategy.catboost import CatBoostClassificationStrategy
from blocksnet.analysis.network.classification.core import NetworkClassifier

strategy = CatBoostClassificationStrategy({
    'iterations':200,
    'learning_rate':0.1,
    'depth':4,
})
classifier = NetworkClassifier(strategy)

[3]:

classifier.prepare_train(graphs)

2025-08-03 22:56:47.286 | INFO     | blocksnet.analysis.network.classification.core:_get_features_df:33 - Preprocessing graphs.
100%|██████████| 355/355 [00:02<00:00, 137.55it/s]
2025-08-03 22:56:49.870 | INFO     | blocksnet.analysis.network.classification.core:_get_features_df:36 - Calculating graphs features.
100%|██████████| 355/355 [01:08<00:00,  5.18it/s]
2025-08-03 22:57:58.426 | SUCCESS  | blocksnet.analysis.network.classification.core:_get_features_df:43 - Features are successfully built.

[4]:

classifier.train()

[4]:

0.7887323943661971

[5]:

# strategy.save('artifacts')

Default strategy

[6]:

from blocksnet.analysis.network.classification.core import NetworkClassifier

classifier = NetworkClassifier.default()

[7]:

classifier.run(graphs)

2025-08-03 22:57:58.689 | INFO     | blocksnet.analysis.network.classification.core:_get_features_df:33 - Preprocessing graphs.
100%|██████████| 355/355 [00:03<00:00, 112.52it/s]
2025-08-03 22:58:01.847 | INFO     | blocksnet.analysis.network.classification.core:_get_features_df:36 - Calculating graphs features.
100%|██████████| 355/355 [01:08<00:00,  5.15it/s]
2025-08-03 22:59:10.842 | SUCCESS  | blocksnet.analysis.network.classification.core:_get_features_df:43 - Features are successfully built.

[7]:

	avg_degree	leaf_nodes_proportion	link_density	avg_clustering	density	diameter	assortativity	avg_near_neighbors_dist	avg_edge_length	ratio_nodes_to_link_length	link_length_entropy	category	town	suburb	village
0	2.632479	0.230769	0.000019	0.044634	0.007521	37	-0.031225	97.796244	201.398576	0.003772	2.379788	SettlementCategory.TOWN	0.966451	0.008857	0.024693
1	2.667722	0.232595	0.000024	0.051160	0.004228	53	0.027810	81.161861	166.624082	0.004499	2.567573	SettlementCategory.TOWN	0.978913	0.009246	0.011841
2	3.127436	0.103448	0.000031	0.034533	0.004696	46	0.165620	92.797600	164.138572	0.003896	2.831230	SettlementCategory.TOWN	0.927734	0.038700	0.033567
3	2.868026	0.162554	0.000045	0.036892	0.001539	80	0.119530	64.962786	139.840803	0.004987	3.061517	SettlementCategory.TOWN	0.958196	0.021080	0.020724
4	2.950000	0.166667	0.000035	0.025000	0.024790	23	0.229978	107.783917	197.933830	0.003425	2.096735	SettlementCategory.TOWN	0.624834	0.051260	0.323906
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
350	2.508475	0.293785	0.000051	0.022599	0.014253	26	-0.112127	67.816268	129.970944	0.006134	2.145329	SettlementCategory.TOWN	0.549521	0.041179	0.409300
351	2.631579	0.175439	0.000061	0.093567	0.023288	26	-0.046066	70.331983	133.376494	0.005698	2.058882	SettlementCategory.VILLAGE	0.385718	0.204624	0.409657
352	2.862745	0.156863	0.000042	0.008170	0.028344	16	0.046019	100.848609	181.909175	0.003841	2.256689	SettlementCategory.VILLAGE	0.360941	0.066698	0.572360
353	2.666667	0.140351	0.000015	0.084795	0.047619	21	-0.002749	128.967290	272.012219	0.002757	1.704535	SettlementCategory.SUBURB	0.207030	0.652064	0.140906
354	2.774566	0.202312	0.000028	0.026975	0.016131	30	0.111970	101.533420	199.022313	0.003622	1.868175	SettlementCategory.TOWN	0.514445	0.107683	0.377872

355 rows × 15 columns