In [1]:

import pandas as pd
import geopandas as gpd
from shapely.geometry import Polygon
from pytorch_lightning import seed_everything

from srai.embedders import GTFS2VecEmbedder
from srai.constants import REGIONS_INDEX

import pandas as pd import geopandas as gpd from shapely.geometry import Polygon from pytorch_lightning import seed_everything from srai.embedders import GTFS2VecEmbedder from srai.constants import REGIONS_INDEX

Example on artificial data¶

Define features and regions¶

In [2]:

features_gdf = gpd.GeoDataFrame(
    {
        "trip_count_at_6": [1, 0, 0],
        "trip_count_at_7": [1, 1, 0],
        "trip_count_at_8": [0, 0, 1],
        "directions_at_6": [
            {"A", "A1"},
            {"B", "B1"},
            {"C"},
        ],
    },
    geometry=gpd.points_from_xy([1, 2, 5], [1, 2, 2]),
    index=pd.Index(name="stop_id", data=[1, 2, 3]),
)
features_gdf

features_gdf = gpd.GeoDataFrame( { "trip_count_at_6": [1, 0, 0], "trip_count_at_7": [1, 1, 0], "trip_count_at_8": [0, 0, 1], "directions_at_6": [ {"A", "A1"}, {"B", "B1"}, {"C"}, ], }, geometry=gpd.points_from_xy([1, 2, 5], [1, 2, 2]), index=pd.Index(name="stop_id", data=[1, 2, 3]), ) features_gdf

Out[2]:

	trip_count_at_6	trip_count_at_7	trip_count_at_8	directions_at_6	geometry
stop_id
1	1	1	0	{A, A1}	POINT (1.00000 1.00000)
2	0	1	0	{B, B1}	POINT (2.00000 2.00000)
3	0	0	1	{C}	POINT (5.00000 2.00000)

In [3]:

regions_gdf = gpd.GeoDataFrame(
    geometry=[
        Polygon([(0, 0), (0, 3), (3, 3), (3, 0)]),
        Polygon([(4, 0), (4, 3), (7, 3), (7, 0)]),
        Polygon([(8, 0), (8, 3), (11, 3), (11, 0)]),
    ],
    index=pd.Index(name=REGIONS_INDEX, data=["ff1", "ff2", "ff3"]),
)
regions_gdf

regions_gdf = gpd.GeoDataFrame( geometry=[ Polygon([(0, 0), (0, 3), (3, 3), (3, 0)]), Polygon([(4, 0), (4, 3), (7, 3), (7, 0)]), Polygon([(8, 0), (8, 3), (11, 3), (11, 0)]), ], index=pd.Index(name=REGIONS_INDEX, data=["ff1", "ff2", "ff3"]), ) regions_gdf

Out[3]:

	geometry
region_id
ff1	POLYGON ((0.00000 0.00000, 0.00000 3.00000, 3....
ff2	POLYGON ((4.00000 0.00000, 4.00000 3.00000, 7....
ff3	POLYGON ((8.00000 0.00000, 8.00000 3.00000, 11...

In [4]:

ax = regions_gdf.plot()
features_gdf.plot(ax=ax, color="red")

ax = regions_gdf.plot() features_gdf.plot(ax=ax, color="red")

Out[4]:

<Axes: >

In [5]:

joint_gdf = gpd.GeoDataFrame()
joint_gdf.index = pd.MultiIndex.from_tuples(
    [("ff1", 1), ("ff1", 2), ("ff2", 3)],
    names=[REGIONS_INDEX, "stop_id"],
)
joint_gdf

joint_gdf = gpd.GeoDataFrame() joint_gdf.index = pd.MultiIndex.from_tuples( [("ff1", 1), ("ff1", 2), ("ff2", 3)], names=[REGIONS_INDEX, "stop_id"], ) joint_gdf

Out[5]:

region_id	stop_id
ff1	1
	2
ff2	3

Get features without embedding them¶

In [6]:

embedder = GTFS2VecEmbedder(skip_autoencoder=True)
res = embedder.transform(regions_gdf, features_gdf, joint_gdf)
res

embedder = GTFS2VecEmbedder(skip_autoencoder=True) res = embedder.transform(regions_gdf, features_gdf, joint_gdf) res

Out[6]:

	directions_at_6
region_id
ff1	1.00
ff2	0.25
ff3	0.00

Fit and train the embedder¶

In [7]:

seed_everything(42)
embedder = GTFS2VecEmbedder(hidden_size=2, embedding_size=4)
embedder.fit(regions_gdf, features_gdf, joint_gdf)
res = embedder.transform(regions_gdf, features_gdf, joint_gdf)
res

seed_everything(42) embedder = GTFS2VecEmbedder(hidden_size=2, embedding_size=4) embedder.fit(regions_gdf, features_gdf, joint_gdf) res = embedder.transform(regions_gdf, features_gdf, joint_gdf) res

Seed set to 42
GPU available: False, used: False
TPU available: False, using: 0 TPU cores
IPU available: False, using: 0 IPUs
HPU available: False, using: 0 HPUs
/opt/hostedtoolcache/Python/3.10.13/x64/lib/python3.10/site-packages/pytorch_lightning/trainer/connectors/logger_connector/logger_connector.py:67: Starting from v1.9.0, `tensorboardX` has been removed as a dependency of the `pytorch_lightning` package, due to potential conflicts with other packages in the ML ecosystem. For this reason, `logger=True` will use `CSVLogger` as the default logger, unless the `tensorboard` or `tensorboardX` packages are found. Please `pip install lightning[extra]` or one of them to enable TensorBoard support by default
Missing logger folder: /home/runner/work/srai/srai/examples/embedders/lightning_logs

  | Name    | Type       | Params
---------------------------------------
0 | encoder | Sequential | 16    
1 | decoder | Sequential | 13    
---------------------------------------
29        Trainable params
0         Non-trainable params
29        Total params
0.000     Total estimated model params size (MB)
/opt/hostedtoolcache/Python/3.10.13/x64/lib/python3.10/site-packages/pytorch_lightning/loops/fit_loop.py:293: The number of training batches (1) is smaller than the logging interval Trainer(log_every_n_steps=50). Set a lower value for log_every_n_steps if you want to see logs for the training epoch.

`Trainer.fit` stopped: `max_epochs=10` reached.

Out[7]:

	0	1	2	3
region_id
ff1	0.657301	0.599207	-0.188990	0.438122
ff2	0.663876	0.541362	-0.220063	0.030094
ff3	0.636288	0.457780	-0.115227	0.004630