Synthetic data with SDV and CopulaGAN

Updated 2024-03-06 (0533e64)

import pandas as pd
import warnings

warnings.filterwarnings('ignore')

data = pd.read_csv("data/svm-hyperparameters-train-features.csv")

data.head()

	Pclass	Sex	Age	SibSp	Fare
0	3	1	22.0	1	7.2500
1	1	0	38.0	1	71.2833
2	3	0	26.0	0	7.9250
3	1	0	35.0	1	53.1000
4	3	1	35.0	0	8.0500

data.describe(include='all')

	Pclass	Sex	Age	SibSp	Parch	Fare
count	891.000000	891.000000	891.000000	891.000000	891.000000	891.000000
mean	2.308642	0.647587	29.758889	0.523008	0.381594	32.204208
std	0.836071	0.477990	13.002570	1.102743	0.806057	49.693429
min	1.000000	0.000000	0.420000	0.000000	0.000000	0.000000
25%	2.000000	0.000000	22.000000	0.000000	0.000000	7.910400
50%	3.000000	1.000000	30.000000	0.000000	0.000000	14.454200
75%	3.000000	1.000000	35.000000	1.000000	0.000000	31.000000
max	3.000000	1.000000	80.000000	8.000000	6.000000	512.329200

from sdv.tabular import CopulaGAN

model = CopulaGAN()

model.fit(data)

new_data = model.sample(200)

new_data.head()

	Pclass	Sex	Age	SibSp	Parch	Fare
0	3	0	32.02	0	0	18.2784
1	3	1	20.90	5	0	109.6910
2	3	0	40.80	0	1	179.5139
3	3	0	33.13	1	0	17.3447
4	1	1	20.62	0	0	9.6040

new_data.describe(include='all')

	Pclass	Sex	Age	SibSp	Parch	Fare
count	200.000000	200.000000	200.000000	200.000000	200.000000	200.000000
mean	2.490000	0.450000	31.404700	0.405000	0.115000	46.756109
std	0.820528	0.498742	12.297721	0.919348	0.461345	52.828023
min	1.000000	0.000000	0.450000	0.000000	0.000000	3.034900
25%	2.000000	0.000000	24.712500	0.000000	0.000000	11.768475
50%	3.000000	0.000000	30.295000	0.000000	0.000000	27.859150
75%	3.000000	1.000000	34.307500	1.000000	0.000000	58.361350
max	3.000000	1.000000	73.340000	7.000000	3.000000	381.859600

from sdv.evaluation import evaluate

evaluate(new_data, data)

0.539229052965023

model = CopulaGAN(
    field_transformers={
        'Pclass': 'categorical',
        'Sex': 'categorical',
        'Age': 'float',
        'SibSp': 'boolean',
        'Parch': 'integer',
        'Fare': 'float'
    },
    field_distributions={
        'Fare': 'truncated_gaussian'
    }
)

model.fit(data)

new_data = model.sample(200)

new_data.head()

	Pclass	Sex	Age	SibSp	Fare
0	1	0	1.07	1	13.8318
1	3	0	15.58	0	46.6937
2	1	1	26.53	0	53.8841
3	1	0	29.58	0	5.2646
4	1	0	30.13	0	5.1415

new_data.describe(include='all')

	Pclass	Sex	Age	SibSp	Parch	Fare
count	200.000000	200.000000	200.000000	200.000000	200.000000	200.000000
mean	2.210000	0.410000	22.935600	0.550000	0.145000	49.155439
std	0.932873	0.493068	13.955182	0.498742	0.441531	42.733997
min	1.000000	0.000000	0.430000	0.000000	0.000000	4.749400
25%	1.000000	0.000000	13.427500	0.000000	0.000000	13.340950
50%	3.000000	0.000000	25.875000	1.000000	0.000000	34.373700
75%	3.000000	1.000000	30.225000	1.000000	0.000000	75.206275
max	3.000000	1.000000	71.340000	1.000000	3.000000	220.761200

evaluate(new_data, data)

0.4983713994365315

Contents

Related

Synthetic data with SDV and CopulaGAN