Service optimization

import pandas as pd

blocks = pd.read_pickle('./../../data/blocks.pickle')
columns = [c for c in blocks.columns if 'capacity_' in c]
blocks = blocks[['geometry', 'population', 'site_area', *columns]].copy()
blocks.head(3)

	geometry	population	site_area	capacity_school	capacity_kindergarten	capacity_hospital	capacity_polyclinic	capacity_pitch	capacity_swimming_pool	capacity_theatre	...	capacity_train_station	capacity_bus_station	capacity_bus_stop	capacity_pier	capacity_animal_shelter	capacity_prison	capacity_landfill	capacity_plant_nursery	capacity_greenhouse_complex	capacity_warehouse
0	POLYGON ((349424.859 6631180.891, 349424.751 6...	0.0	0.032316	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
1	POLYGON ((352083.617 6633950.146, 352240.448 6...	0.0	137101.088247	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	20.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
2	POLYGON ((346700.642 6618453.176, 346681.107 6...	109.0	95239.084434	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0

3 rows × 63 columns

accessibility_matrix = pd.read_pickle('./../../data/acc_mx.pickle')
accessibility_matrix.head()

/home/ivan/work/ovanes-projects/urban-planning-optimizer/general/.venv/lib/python3.12/site-packages/pandas/io/formats/format.py:1458: RuntimeWarning: overflow encountered in cast
  has_large_values = (abs_vals > 1e6).any()

	0	1	2	3	4	5	6	7	8	9	...	9358	9359	9360	9361	9362	9363	9364	9365	9366	9367
0	0.0000	58.15625	74.187500	76.937500	66.8125	80.5000	92.18750	66.56250	65.68750	106.8750	...	70.0000	72.1875	72.0000	73.3125	73.2500	74.0000	75.5000	83.1875	75.0000	73.2500
1	56.2500	0.00000	99.687500	102.375000	88.1250	76.1250	113.43750	89.43750	88.56250	96.3125	...	65.6250	67.7500	67.6875	69.0000	68.8750	69.6875	71.1875	78.8125	70.6875	68.8750
2	74.5000	96.50000	0.000000	5.347656	74.0000	118.8125	92.06250	64.43750	63.56250	134.5000	...	108.3125	110.4375	110.3750	111.6875	111.5625	112.3750	113.8125	121.5000	113.3750	111.5625
3	77.2500	99.25000	5.347656	0.000000	76.7500	121.5625	90.12500	67.18750	66.31250	137.2500	...	111.0625	113.1875	113.0625	114.3750	114.3125	115.0625	116.5625	124.2500	116.1250	114.3125
4	70.1875	90.56250	82.375000	85.125000	0.0000	110.5000	35.53125	45.15625	44.28125	89.8750	...	100.0000	102.1250	102.0000	103.3125	103.2500	104.0000	105.5000	113.1875	105.0000	103.2500

5 rows × 9368 columns

Select ID of optimized blocks

from blocksnet.enums import LandUse
blocks_lu = {
    2059 : LandUse.RESIDENTIAL
}

Service distribution

basic_services = {
    "kindergarten": 0.2,
    "school": 0.2,
    "pharmacy": 0.2,
    "polyclinics": 0.3,
    "convenience": 0.2,
    "cafe": 0.15,
    "playground": 0.2,
    "post": 0.1,
    "hairdresser": 0.1
}
advanced_services = {
    "hospital": 0.5,
    "supermarket": 0.4,
    "restaurant": 0.3,
    "park": 0.8,
    "bank": 0.3,
    "fuel": 0.25,
    "police": 0.2,
    "cinema": 0.2,
    "market": 0.2,
    "religion": 0.5,
    "multifunctional_center": 0.1,
    "train_building": 0.1,
    "cemetery": 0.1,
    "notary": 0.1,
    "pitch": 0.2
}
comfort_services = {
    "mall": 0.3,
    "swimming_pool": 0.4,
    "theatre": 0.2,
    "museum": 0.2,
    "university": 0.2,
    "sanatorium": 0.2,
    "zoo": 0.1,
    "reserve": 0.1,
    "animal_shelter": 0.1,
    "aeroway_terminal": 0.2,
    "crematorium": 0.1,
    "prison": 0.1,
    "landfill": 0.1,
    "bar": 0.3,
    "hotel": 0.3,
}


def get_service_weight(service_name):
    if service_name in basic_services:
        return 0.5714 * basic_services[service_name]
    elif service_name in advanced_services:
        return 0.2857 * advanced_services[service_name]
    elif service_name in comfort_services:
        return 0.1429 * comfort_services[service_name]
    return 0

Service types and weights

chosen_service_types = set(basic_services) | set(advanced_services) | set(comfort_services)
service_weights = {service_type: get_service_weight(service_type) for service_type in chosen_service_types if f'capacity_{service_type}' in blocks.columns}

Initialize and run optimizer

from blocksnet.optimization.services import (
    TPEOptimizer,
    WeightedObjective,
    WeightedConstraints,
    Facade,
    BlockSolution,
    GradientChooser,
    SimpleChooser,
    RandomOrder,
    AreaSolution
)
from tqdm import tqdm

var_adapter = AreaSolution(blocks_lu)

facade = Facade(
    blocks_lu=blocks_lu,
    blocks_df=blocks,
    accessibility_matrix=accessibility_matrix,
    var_adapter=var_adapter,
)

for service_type, weight in tqdm(service_weights.items()):
    facade.add_service_type(service_type, weight, blocks.rename(columns={f'capacity_{service_type}': 'capacity'})[['capacity']])

objective = WeightedObjective(num_params=facade.num_params, facade=facade, weights=service_weights, max_evals=100)

constraints = WeightedConstraints(num_params=facade.num_params, facade=facade, priority=service_weights)

tpe_optimizer = TPEOptimizer(
    objective=objective, constraints=constraints, vars_order=RandomOrder(), vars_chooser=GradientChooser(facade, num_params=facade.num_params, num_top=5)
)

best_x, best_val, perc, func_evals = tpe_optimizer.run(max_runs=1000, timeout=60000, initial_runs_num=1)

100%|██████████| 35/35 [13:15<00:00, 22.74s/it]
[I 2025-07-30 13:47:33,551] A new study created in memory with name: no-name-7236c4ec-218f-472c-9099-73a2aff3b9ff

[I 2025-07-30 13:47:36,172] Trial 0 finished with value: 1.7590525543245839 and parameters: {'x_12': 0, 'x_6': 700, 'x_5': 60, 'x_14': 10000, 'x_15': 0, 'x_24': 0, 'x_4': 400, 'x_19': 0, 'x_13': 50, 'x_23': 0, 'x_20': 0, 'x_11': 0, 'x_21': 100, 'x_17': 9100, 'x_1': 1000, 'x_9': 20000, 'x_2': 450, 'x_7': 270, 'x_25': 150, 'x_18': 2800, 'x_16': 0, 'x_0': 30, 'x_22': 3520, 'x_3': 18500, 'x_10': 0, 'x_8': 0}. Best is trial 0 with value: 1.7590525543245839.
[I 2025-07-30 13:47:38,178] Trial 1 finished with value: 1.7591407549631215 and parameters: {'x_17': 9100, 'x_18': 2802, 'x_9': 20001, 'x_8': 8, 'x_12': 0, 'x_4': 400, 'x_0': 30, 'x_19': 9, 'x_25': 151, 'x_6': 700, 'x_7': 276, 'x_13': 50, 'x_15': 3, 'x_1': 1000, 'x_11': 5, 'x_21': 100, 'x_14': 10006, 'x_20': 1, 'x_10': 0, 'x_3': 18512, 'x_24': 6, 'x_2': 450, 'x_16': 6, 'x_5': 61, 'x_23': 5, 'x_22': 3533}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:40,099] Trial 2 finished with value: 1.7538942265405484 and parameters: {'x_9': 123, 'x_11': 15893, 'x_2': 117, 'x_18': 3990, 'x_0': 22, 'x_22': 3520, 'x_23': 0, 'x_10': 0, 'x_19': 0, 'x_1': 1000, 'x_14': 10000, 'x_16': 0, 'x_21': 100, 'x_7': 270, 'x_13': 50, 'x_6': 700, 'x_25': 150, 'x_17': 9100, 'x_24': 0, 'x_5': 60, 'x_4': 400, 'x_8': 0, 'x_15': 0, 'x_3': 0, 'x_20': 0, 'x_12': 0}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:41,837] Trial 3 finished with value: 1.7538942265405484 and parameters: {'x_2': 519, 'x_15': 25045, 'x_20': 16569, 'x_5': 77, 'x_24': 0, 'x_10': 0, 'x_25': 150, 'x_7': 270, 'x_18': 2800, 'x_23': 0, 'x_0': 0, 'x_16': 0, 'x_11': 0, 'x_6': 700, 'x_9': 0, 'x_4': 400, 'x_3': 0, 'x_21': 100, 'x_13': 50, 'x_14': 10000, 'x_1': 1000, 'x_19': 0, 'x_22': 3520, 'x_12': 0, 'x_8': 0, 'x_17': 9100}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:42,582] Trial 4 finished with value: 1.6321306966077258 and parameters: {'x_16': 70111, 'x_2': 112, 'x_15': 55, 'x_7': 21, 'x_22': 0, 'x_12': 0, 'x_23': 0, 'x_19': 0, 'x_10': 0, 'x_21': 100, 'x_25': 150, 'x_9': 0, 'x_1': 1000, 'x_4': 400, 'x_5': 60, 'x_3': 0, 'x_18': 2800, 'x_24': 0, 'x_13': 50, 'x_11': 0, 'x_8': 0, 'x_14': 0, 'x_6': 0, 'x_20': 0, 'x_0': 0, 'x_17': 0}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:42,912] Trial 5 finished with value: 1.6321306966077258 and parameters: {'x_20': 48633, 'x_6': 102, 'x_18': 2804, 'x_15': 0, 'x_13': 50, 'x_2': 0, 'x_1': 1001, 'x_14': 0, 'x_12': 0, 'x_19': 0, 'x_22': 0, 'x_21': 100, 'x_9': 0, 'x_4': 400, 'x_8': 0, 'x_5': 60, 'x_7': 0, 'x_11': 0, 'x_25': 150, 'x_3': 0, 'x_23': 0, 'x_0': 0, 'x_24': 0, 'x_17': 0, 'x_10': 0, 'x_16': 17000}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:43,386] Trial 6 finished with value: 1.7304942620673358 and parameters: {'x_20': 70126, 'x_0': 32, 'x_2': 60, 'x_12': 82, 'x_23': 0, 'x_10': 0, 'x_3': 0, 'x_9': 0, 'x_24': 0, 'x_17': 0, 'x_7': 0, 'x_16': 17000, 'x_22': 0, 'x_5': 60, 'x_18': 2800, 'x_14': 0, 'x_25': 150, 'x_15': 0, 'x_13': 50, 'x_4': 400, 'x_11': 0, 'x_8': 0, 'x_6': 0, 'x_1': 0, 'x_21': 0, 'x_19': 0}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:43,778] Trial 7 finished with value: 1.7304942620673358 and parameters: {'x_8': 160, 'x_1': 551, 'x_10': 992, 'x_11': 47921, 'x_18': 2882, 'x_0': 31, 'x_17': 102, 'x_21': 0, 'x_12': 1, 'x_9': 0, 'x_7': 0, 'x_24': 0, 'x_5': 60, 'x_20': 0, 'x_22': 0, 'x_3': 0, 'x_14': 0, 'x_6': 0, 'x_19': 0, 'x_25': 150, 'x_13': 50, 'x_23': 0, 'x_4': 400, 'x_16': 17000, 'x_15': 0, 'x_2': 0}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:44,293] Trial 8 finished with value: 1.731829308796308 and parameters: {'x_25': 16, 'x_2': 308, 'x_7': 1189, 'x_21': 61, 'x_4': 7, 'x_12': 0, 'x_23': 0, 'x_18': 2800, 'x_15': 0, 'x_19': 0, 'x_5': 60, 'x_6': 0, 'x_24': 0, 'x_16': 17000, 'x_0': 30, 'x_1': 0, 'x_3': 0, 'x_13': 50, 'x_17': 0, 'x_11': 0, 'x_22': 0, 'x_8': 0, 'x_10': 0, 'x_20': 0, 'x_9': 0, 'x_14': 0}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:44,607] Trial 9 finished with value: 1.731829308796308 and parameters: {'x_9': 19915, 'x_7': 1460, 'x_12': 28608, 'x_23': 176, 'x_11': 89, 'x_14': 111, 'x_5': 61, 'x_6': 0, 'x_2': 0, 'x_16': 17000, 'x_18': 2800, 'x_25': 0, 'x_22': 0, 'x_4': 0, 'x_15': 0, 'x_20': 0, 'x_10': 0, 'x_21': 0, 'x_3': 0, 'x_1': 0, 'x_8': 0, 'x_19': 0, 'x_17': 0, 'x_13': 50, 'x_0': 30, 'x_24': 0}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:45,114] Trial 10 finished with value: 1.731829308796308 and parameters: {'x_19': 1608, 'x_21': 67, 'x_0': 12, 'x_15': 8173, 'x_24': 21918, 'x_2': 0, 'x_23': 0, 'x_22': 0, 'x_17': 0, 'x_8': 0, 'x_7': 1350, 'x_14': 0, 'x_3': 0, 'x_4': 0, 'x_1': 0, 'x_16': 17000, 'x_11': 0, 'x_20': 0, 'x_13': 50, 'x_18': 2800, 'x_10': 0, 'x_25': 0, 'x_6': 0, 'x_12': 0, 'x_9': 0, 'x_5': 0}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:45,656] Trial 11 finished with value: 1.731829308796308 and parameters: {'x_12': 14916, 'x_22': 2692, 'x_3': 18538, 'x_17': 6892, 'x_18': 3169, 'x_21': 77, 'x_4': 282, 'x_24': 1653, 'x_15': 1103, 'x_5': 105, 'x_11': 765, 'x_13': 50, 'x_20': 540, 'x_8': 92, 'x_23': 512, 'x_1': 526, 'x_6': 38, 'x_10': 0, 'x_16': 17000, 'x_14': 0, 'x_25': 0, 'x_2': 0, 'x_9': 0, 'x_7': 1350, 'x_0': 0, 'x_19': 0}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:46,145] Trial 12 finished with value: 1.731829308796308 and parameters: {'x_23': 140, 'x_16': 49876, 'x_22': 2635, 'x_14': 7698, 'x_10': 293, 'x_6': 0, 'x_20': 0, 'x_1': 0, 'x_7': 1350, 'x_13': 50, 'x_4': 0, 'x_3': 18500, 'x_2': 0, 'x_9': 0, 'x_25': 0, 'x_0': 0, 'x_21': 0, 'x_8': 0, 'x_19': 0, 'x_11': 0, 'x_18': 2800, 'x_15': 0, 'x_24': 0, 'x_5': 0, 'x_12': 0, 'x_17': 0}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:46,683] Trial 13 finished with value: 1.731829308796308 and parameters: {'x_23': 377, 'x_19': 958, 'x_5': 29, 'x_8': 81, 'x_11': 7762, 'x_6': 548, 'x_0': 23, 'x_2': 448, 'x_17': 6880, 'x_16': 20640, 'x_25': 200, 'x_13': 50, 'x_9': 9646, 'x_1': 7, 'x_21': 1, 'x_10': 0, 'x_20': 0, 'x_3': 18500, 'x_7': 1350, 'x_4': 0, 'x_14': 0, 'x_12': 0, 'x_18': 2800, 'x_22': 0, 'x_24': 0, 'x_15': 0}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:47,275] Trial 14 finished with value: 1.731829308796308 and parameters: {'x_5': 35, 'x_17': 7322, 'x_3': 14419, 'x_24': 11194, 'x_7': 662, 'x_19': 0, 'x_12': 0, 'x_10': 0, 'x_15': 0, 'x_20': 0, 'x_1': 0, 'x_13': 50, 'x_16': 17000, 'x_0': 0, 'x_2': 0, 'x_8': 0, 'x_18': 2800, 'x_22': 0, 'x_6': 0, 'x_11': 0, 'x_4': 0, 'x_21': 0, 'x_14': 0, 'x_23': 0, 'x_9': 0, 'x_25': 0}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:47,719] Trial 15 finished with value: 1.731829308796308 and parameters: {'x_16': 28408, 'x_10': 244, 'x_15': 5851, 'x_17': 3535, 'x_3': 13993, 'x_19': 3480, 'x_8': 297, 'x_0': 24, 'x_7': 873, 'x_20': 2591, 'x_9': 8146, 'x_13': 50, 'x_1': 7, 'x_12': 1, 'x_6': 0, 'x_4': 0, 'x_5': 0, 'x_18': 2800, 'x_14': 0, 'x_23': 0, 'x_2': 0, 'x_11': 0, 'x_21': 0, 'x_25': 0, 'x_24': 0, 'x_22': 0}. Best is trial 1 with value: 1.7591407549631215.
[I 2025-07-30 13:47:49,809] Trial 16 finished with value: 1.8061746636207854 and parameters: {'x_2': 335, 'x_9': 19998, 'x_20': 11550, 'x_3': 5993, 'x_15': 16670, 'x_8': 0, 'x_19': 0, 'x_6': 0, 'x_23': 0, 'x_1': 0, 'x_17': 0, 'x_13': 50, 'x_14': 0, 'x_11': 0, 'x_0': 0, 'x_25': 0, 'x_5': 0, 'x_21': 0, 'x_18': 2800, 'x_24': 0, 'x_7': 810, 'x_22': 0, 'x_16': 0, 'x_4': 0, 'x_12': 0, 'x_10': 0}. Best is trial 16 with value: 1.8061746636207854.
[I 2025-07-30 13:47:51,973] Trial 17 finished with value: 1.8061746636207854 and parameters: {'x_18': 3089, 'x_25': 561, 'x_15': 18264, 'x_14': 2808, 'x_10': 728, 'x_9': 16344, 'x_21': 240, 'x_17': 2403, 'x_2': 304, 'x_13': 53, 'x_1': 299, 'x_7': 1979, 'x_23': 1156, 'x_12': 4818, 'x_3': 4874, 'x_20': 6314, 'x_4': 1045, 'x_5': 238, 'x_22': 1093, 'x_16': 713, 'x_24': 662, 'x_8': 849, 'x_6': 232, 'x_19': 266, 'x_11': 243, 'x_0': 100}. Best is trial 16 with value: 1.8061746636207854.
[I 2025-07-30 13:47:54,123] Trial 18 finished with value: 1.7158014038955691 and parameters: {'x_22': 757, 'x_13': 39, 'x_24': 42661, 'x_10': 774, 'x_17': 1916, 'x_15': 0, 'x_23': 0, 'x_16': 0, 'x_20': 0, 'x_7': 1350, 'x_12': 0, 'x_4': 400, 'x_21': 100, 'x_6': 0, 'x_1': 0, 'x_18': 2800, 'x_25': 150, 'x_3': 0, 'x_0': 30, 'x_9': 0, 'x_8': 0, 'x_19': 0, 'x_14': 0, 'x_5': 0, 'x_11': 0, 'x_2': 0}. Best is trial 16 with value: 1.8061746636207854.
[I 2025-07-30 13:47:56,131] Trial 19 finished with value: 1.7158014038955691 and parameters: {'x_18': 3066, 'x_9': 16515, 'x_23': 1700, 'x_3': 4527, 'x_12': 2777, 'x_22': 1438, 'x_1': 238, 'x_7': 1799, 'x_19': 738, 'x_0': 135, 'x_14': 1550, 'x_21': 208, 'x_5': 199, 'x_8': 599, 'x_2': 273, 'x_20': 2555, 'x_10': 623, 'x_13': 13, 'x_15': 790, 'x_16': 2, 'x_24': 30003, 'x_6': 1, 'x_4': 401, 'x_11': 0, 'x_25': 150, 'x_17': 0}. Best is trial 16 with value: 1.8061746636207854.

Convert to services dataframe

solution_df = facade.solution_to_services_df(best_x)
solution_df.to_csv(f'services.csv')
solution_df.head(5)

	block_id	service_type	site_area	build_floor_area	capacity	count
3	2059	convenience	0.0	25.0	50	1
8	2059	cinema	0.0	700.0	450	1
12	2059	bank	0.0	400.0	250	1
18	2059	school	0.0	2200.0	250	1
27	2059	bar	0.0	40.0	30	1

Extract area dataframe from solution

solution_area_df = facade.get_solution_area_df(best_x)
solution_area_df.to_csv(f'solution_area.csv')
solution_area_df.loc[blocks_lu.keys()].head(5)

	population	build_floor_area	living_area	footprint_area
2059	2631.0	79087.734512	28192.734512	14060.041691

Visualization

import matplotlib.pyplot as plt


def plot_df(df):
    df_called_obj = df.loc[df["called_obj"] == True]

    plt.scatter(df_called_obj["func_evals"], df_called_obj["best_val"], c="red", s=5)

    df_called_obj = pd.concat(
        [pd.DataFrame([[-1, None, 0, 0, None, 0, False, 0]], columns=df_called_obj.columns), df_called_obj]
    )
    plt.plot(
        df_called_obj["func_evals"], df_called_obj["best_val"], label="Best Value (TPE)", color="blue", linewidth=1
    )
    # Добавление точек с цветом в зависимости от 'called_obj'
    # Настройка осей и заголовка
    plt.xlabel("Number of objective evaluation")
    plt.ylabel("Best Value")
    plt.title("Best Value Across Trials")
    plt.legend()
    plt.grid(True)

    # Отображение графика
    plt.show()

import pandas as pd

tpe_df = pd.read_csv("tpe_trials.csv")
plot_df(tpe_df)

facade.save_delta_demand(best_x)
population_df = pd.read_csv("demand_services.csv")
population_df.head(20)

	Block id	Service type	Demand
0	2059	convenience	473
1	2059	cinema	21
2	2059	bank	52
3	2059	school	315
4	2059	bar	65
5	2059	fuel	78
6	2059	police	13
7	2059	restaurant	186
8	2059	hairdresser	26
9	2059	university	34
10	2059	museum	7
11	2059	multifunctional_center	210
12	2059	pharmacy	39
13	2059	playground	5
14	2059	mall	21
15	2059	hotel	23
16	2059	market	394
17	2059	religion	26
18	2059	theatre	15
19	2059	hospital	23