Examples

Credit Scoring Example

Settings¶

Installation¶

In [1]:

!pip install --upgrade autocarver

Requirement already satisfied: autocarver in c:\users\defra\.conda\envs\py39\lib\site-packages (5.1.2)
Requirement already satisfied: ipython in c:\users\defra\.conda\envs\py39\lib\site-packages (from autocarver) (8.11.0)
Requirement already satisfied: numpy in c:\users\defra\.conda\envs\py39\lib\site-packages (from autocarver) (1.24.2)
Requirement already satisfied: tqdm in c:\users\defra\.conda\envs\py39\lib\site-packages (from autocarver) (4.65.0)
Requirement already satisfied: scipy in c:\users\defra\.conda\envs\py39\lib\site-packages (from autocarver) (1.10.1)
Requirement already satisfied: statsmodels in c:\users\defra\.conda\envs\py39\lib\site-packages (from autocarver) (0.14.0)
Requirement already satisfied: pandas in c:\users\defra\.conda\envs\py39\lib\site-packages (from autocarver) (1.5.3)
Requirement already satisfied: scikit-learn in c:\users\defra\.conda\envs\py39\lib\site-packages (from autocarver) (1.2.2)
Requirement already satisfied: decorator in c:\users\defra\.conda\envs\py39\lib\site-packages (from ipython->autocarver) (5.1.1)
Requirement already satisfied: jedi>=0.16 in c:\users\defra\.conda\envs\py39\lib\site-packages (from ipython->autocarver) (0.18.2)
Requirement already satisfied: traitlets>=5 in c:\users\defra\.conda\envs\py39\lib\site-packages (from ipython->autocarver) (5.9.0)
Requirement already satisfied: stack-data in c:\users\defra\.conda\envs\py39\lib\site-packages (from ipython->autocarver) (0.6.2)
Requirement already satisfied: matplotlib-inline in c:\users\defra\.conda\envs\py39\lib\site-packages (from ipython->autocarver) (0.1.6)
Requirement already satisfied: colorama in c:\users\defra\.conda\envs\py39\lib\site-packages (from ipython->autocarver) (0.4.6)
Requirement already satisfied: backcall in c:\users\defra\.conda\envs\py39\lib\site-packages (from ipython->autocarver) (0.2.0)
Requirement already satisfied: pickleshare in c:\users\defra\.conda\envs\py39\lib\site-packages (from ipython->autocarver) (0.7.5)
Requirement already satisfied: pygments>=2.4.0 in c:\users\defra\.conda\envs\py39\lib\site-packages (from ipython->autocarver) (2.14.0)
Requirement already satisfied: prompt-toolkit!=3.0.37,<3.1.0,>=3.0.30 in c:\users\defra\.conda\envs\py39\lib\site-packages (from ipython->autocarver) (3.0.38)
Requirement already satisfied: pytz>=2020.1 in c:\users\defra\.conda\envs\py39\lib\site-packages (from pandas->autocarver) (2022.7.1)
Requirement already satisfied: python-dateutil>=2.8.1 in c:\users\defra\.conda\envs\py39\lib\site-packages (from pandas->autocarver) (2.8.2)
Requirement already satisfied: joblib>=1.1.1 in c:\users\defra\.conda\envs\py39\lib\site-packages (from scikit-learn->autocarver) (1.2.0)
Requirement already satisfied: threadpoolctl>=2.0.0 in c:\users\defra\.conda\envs\py39\lib\site-packages (from scikit-learn->autocarver) (3.1.0)
Requirement already satisfied: patsy>=0.5.2 in c:\users\defra\.conda\envs\py39\lib\site-packages (from statsmodels->autocarver) (0.5.3)
Requirement already satisfied: packaging>=21.3 in c:\users\defra\.conda\envs\py39\lib\site-packages (from statsmodels->autocarver) (23.0)
Requirement already satisfied: parso<0.9.0,>=0.8.0 in c:\users\defra\.conda\envs\py39\lib\site-packages (from jedi>=0.16->ipython->autocarver) (0.8.3)
Requirement already satisfied: six in c:\users\defra\.conda\envs\py39\lib\site-packages (from patsy>=0.5.2->statsmodels->autocarver) (1.16.0)
Requirement already satisfied: wcwidth in c:\users\defra\.conda\envs\py39\lib\site-packages (from prompt-toolkit!=3.0.37,<3.1.0,>=3.0.30->ipython->autocarver) (0.2.6)
Requirement already satisfied: pure-eval in c:\users\defra\.conda\envs\py39\lib\site-packages (from stack-data->ipython->autocarver) (0.2.2)
Requirement already satisfied: asttokens>=2.1.0 in c:\users\defra\.conda\envs\py39\lib\site-packages (from stack-data->ipython->autocarver) (2.2.1)
Requirement already satisfied: executing>=1.2.0 in c:\users\defra\.conda\envs\py39\lib\site-packages (from stack-data->ipython->autocarver) (1.2.0)

Setting up samples¶

This dataset can be found from the corresponding Kaggle competition at https://www.kaggle.com/competitions/GiveMeSomeCredit/

In [1]:

import pandas as pd

data_path = "GiveMeSomeCredit"

credit_data = pd.read_csv(f"{data_path}/cs-training.csv", index_col=0)
print(credit_data.shape)
credit_data.head()

(150000, 11)

Out[1]:

	SeriousDlqin2yrs	RevolvingUtilizationOfUnsecuredLines	age	NumberOfTime30-59DaysPastDueNotWorse	DebtRatio	MonthlyIncome	NumberOfOpenCreditLinesAndLoans	NumberOfTimes90DaysLate	NumberRealEstateLoansOrLines	NumberOfTime60-89DaysPastDueNotWorse	NumberOfDependents
1	1	0.766127	45	2	0.802982	9120.0	13	0	6	0	2.0
2	0	0.957151	40	0	0.121876	2600.0	4	0	0	0	1.0
3	0	0.658180	38	1	0.085113	3042.0	2	1	0	0	0.0
4	0	0.233810	30	0	0.036050	3300.0	5	0	0	0	0.0
5	0	0.907239	49	1	0.024926	63588.0	7	0	1	0	0.0

In [2]:

from sklearn.model_selection import train_test_split

X_train, X_dev = train_test_split(credit_data, test_size=0.33, random_state=42)

Picking up columns to Carve¶

In [3]:

X_train.dtypes

Out[3]:

SeriousDlqin2yrs                          int64
RevolvingUtilizationOfUnsecuredLines    float64
age                                       int64
NumberOfTime30-59DaysPastDueNotWorse      int64
DebtRatio                               float64
MonthlyIncome                           float64
NumberOfOpenCreditLinesAndLoans           int64
NumberOfTimes90DaysLate                   int64
NumberRealEstateLoansOrLines              int64
NumberOfTime60-89DaysPastDueNotWorse      int64
NumberOfDependents                      float64
dtype: object

In [4]:

X_train.isna().mean()

Out[4]:

SeriousDlqin2yrs                        0.000000
RevolvingUtilizationOfUnsecuredLines    0.000000
age                                     0.000000
NumberOfTime30-59DaysPastDueNotWorse    0.000000
DebtRatio                               0.000000
MonthlyIncome                           0.197383
NumberOfOpenCreditLinesAndLoans         0.000000
NumberOfTimes90DaysLate                 0.000000
NumberRealEstateLoansOrLines            0.000000
NumberOfTime60-89DaysPastDueNotWorse    0.000000
NumberOfDependents                      0.026129
dtype: float64

In [5]:

target = "SeriousDlqin2yrs"
quantitative_features = [feature for feature in X_train if feature != target]

Feature processing with AutoCarver¶

Fitting train samples and testint robustness¶

In [6]:

from AutoCarver import AutoCarver

auto_carver = AutoCarver(
    quantitative_features=quantitative_features,
    qualitative_features=[],
    sort_by='cramerv',  # Best combination according to Cramer's V
    dropna=False,  # don't want to group nans with other values, leave that to XGBoost 
    min_freq=0.1,  # minimum frequency per modality
    max_n_mod=5,  # maximum number of modality per carved feature
    copy=True,  # in order not to modify X_train directly
    pretty_print=True,  # prints nice tables
)
x_discretized = auto_carver.fit_transform(
    # specifying dataset to carve
    X_train, X_train[target],
    # specifying a dataset to test robustness
    X_dev=X_dev, y_dev=X_dev[target]
)

------
[Discretizer] Fit Quantitative Features
---
 - [QuantileDiscretizer] Fit ['age', 'NumberOfDependents', 'DebtRatio', 'RevolvingUtilizationOfUnsecuredLines', 'NumberOfTimes90DaysLate', 'NumberOfTime60-89DaysPastDueNotWorse', 'NumberOfOpenCreditLinesAndLoans', 'NumberRealEstateLoansOrLines', 'NumberOfTime30-59DaysPastDueNotWorse', 'MonthlyIncome']
 - [BaseDiscretizer] Transform Quantitative ['age', 'NumberOfDependents', 'DebtRatio', 'RevolvingUtilizationOfUnsecuredLines', 'NumberOfTimes90DaysLate', 'NumberOfTime60-89DaysPastDueNotWorse', 'NumberOfOpenCreditLinesAndLoans', 'NumberRealEstateLoansOrLines', 'NumberOfTime30-59DaysPastDueNotWorse', 'MonthlyIncome']
 - [OrdinalDiscretizer] Fit ['NumberOfDependents', 'NumberOfTime30-59DaysPastDueNotWorse', 'DebtRatio', 'RevolvingUtilizationOfUnsecuredLines', 'NumberOfTimes90DaysLate', 'NumberOfTime60-89DaysPastDueNotWorse', 'NumberOfOpenCreditLinesAndLoans', 'NumberRealEstateLoansOrLines', 'age', 'MonthlyIncome']
------

 - [BaseDiscretizer] Transform Quantitative ['NumberOfDependents', 'NumberOfTime30-59DaysPastDueNotWorse', 'DebtRatio', 'RevolvingUtilizationOfUnsecuredLines', 'NumberOfTimes90DaysLate', 'NumberOfTime60-89DaysPastDueNotWorse', 'NumberOfOpenCreditLinesAndLoans', 'NumberRealEstateLoansOrLines', 'age', 'MonthlyIncome']
 - [BaseDiscretizer] Transform Quantitative ['NumberOfDependents', 'NumberOfTime30-59DaysPastDueNotWorse', 'DebtRatio', 'RevolvingUtilizationOfUnsecuredLines', 'NumberOfTimes90DaysLate', 'NumberOfTime60-89DaysPastDueNotWorse', 'NumberOfOpenCreditLinesAndLoans', 'NumberRealEstateLoansOrLines', 'age', 'MonthlyIncome']

------
[AutoCarver] Fit NumberOfDependents (1/10)
---

 - [AutoCarver] Raw feature distribution

X distribution

	target_rate	frequency
NumberOfDependents
x <= 0.0	0.060000	0.580000
0.0 < x <= 1.0	0.073000	0.175000
1.0 < x	0.085000	0.219000
__NAN__	0.048000	0.026000

X_dev distribution

	target_rate	frequency
NumberOfDependents
x <= 0.0	0.057000	0.579000
0.0 < x <= 1.0	0.074000	0.177000
1.0 < x	0.086000	0.218000
__NAN__	0.042000	0.026000

Grouping modalities   : 100%|██████████| 3/3 [00:00<?, ?it/s]
Computing associations: 100%|██████████| 3/3 [00:00<00:00, 3078.76it/s]
Testing robustness    :   0%|          | 0/3 [00:00<?, ?it/s]

 - [AutoCarver] Carved feature distribution

X distribution

	target_rate	frequency
x <= 0.0	0.060000	0.580000
0.0 < x <= 1.0	0.073000	0.175000
1.0 < x	0.085000	0.219000
__NAN__	0.048000	0.026000

X_dev distribution

	target_rate	frequency
x <= 0.0	0.057000	0.579000
0.0 < x <= 1.0	0.074000	0.177000
1.0 < x	0.086000	0.218000
__NAN__	0.042000	0.026000

------


------
[AutoCarver] Fit NumberOfTime30-59DaysPastDueNotWorse (2/10)
---

 - [AutoCarver] Raw feature distribution

X distribution

	target_rate	frequency
NumberOfTime30-59DaysPastDueNotWorse
x <= 0.0	0.040000	0.839000
0.0 < x	0.208000	0.161000

X_dev distribution

	target_rate	frequency
NumberOfTime30-59DaysPastDueNotWorse
x <= 0.0	0.039000	0.842000
0.0 < x	0.207000	0.158000

Grouping modalities   : 100%|██████████| 1/1 [00:00<?, ?it/s]
Computing associations: 100%|██████████| 1/1 [00:00<00:00, 1001.98it/s]
Testing robustness    :   0%|          | 0/1 [00:00<?, ?it/s]

 - [AutoCarver] Carved feature distribution

X distribution

	target_rate	frequency
x <= 0.0	0.040000	0.839000
0.0 < x	0.208000	0.161000

X_dev distribution

	target_rate	frequency
x <= 0.0	0.039000	0.842000
0.0 < x	0.207000	0.158000

------


------
[AutoCarver] Fit DebtRatio (3/10)
---

 - [AutoCarver] Raw feature distribution

X distribution

	target_rate	frequency
DebtRatio
x <= 31.1m	0.052000	0.100000
31.1m < x <= 133.6m	0.070000	0.100000
133.6m < x <= 213.9m	0.061000	0.100000
213.9m < x <= 287.6m	0.054000	0.100000
287.6m < x <= 467.4m	0.061000	0.200000
467.4m < x <= 648.0m	0.088000	0.100000
648.0m < x <= 3.8	0.114000	0.100000
3.8 < x	0.056000	0.200000

X_dev distribution

	target_rate	frequency
DebtRatio
x <= 31.1m	0.056000	0.101000
31.1m < x <= 133.6m	0.064000	0.099000
133.6m < x <= 213.9m	0.058000	0.101000
213.9m < x <= 287.6m	0.055000	0.100000
287.6m < x <= 467.4m	0.062000	0.199000
467.4m < x <= 648.0m	0.077000	0.099000
648.0m < x <= 3.8	0.115000	0.100000
3.8 < x	0.054000	0.202000

Grouping modalities   : 100%|██████████| 98/98 [00:00<00:00, 7372.02it/s]
Computing associations: 100%|██████████| 98/98 [00:00<00:00, 3828.24it/s]
Testing robustness    :   2%|▏         | 2/98 [00:00<00:00, 799.98it/s]

 - [AutoCarver] Carved feature distribution

X distribution

	target_rate	frequency
x <= 31.1m	0.059000	0.400000
287.6m < x <= 467.4m	0.061000	0.200000
467.4m < x <= 648.0m	0.088000	0.100000
648.0m < x <= 3.8	0.114000	0.100000
3.8 < x	0.056000	0.200000

X_dev distribution

	target_rate	frequency
x <= 31.1m	0.059000	0.401000
287.6m < x <= 467.4m	0.062000	0.199000
467.4m < x <= 648.0m	0.077000	0.099000
648.0m < x <= 3.8	0.115000	0.100000
3.8 < x	0.054000	0.202000

------


------
[AutoCarver] Fit RevolvingUtilizationOfUnsecuredLines (4/10)
---

 - [AutoCarver] Raw feature distribution

X distribution

	target_rate	frequency
RevolvingUtilizationOfUnsecuredLines
x <= 3.0m	0.025000	0.100000
3.0m < x <= 19.2m	0.014000	0.100000
19.2m < x <= 43.6m	0.014000	0.100000
43.6m < x <= 83.8m	0.019000	0.100000
83.8m < x <= 155.4m	0.025000	0.100000
155.4m < x <= 273.6m	0.037000	0.100000
273.6m < x <= 447.4m	0.053000	0.100000
447.4m < x <= 700.7m	0.088000	0.100000
700.7m < x	0.199000	0.200000

X_dev distribution

	target_rate	frequency
RevolvingUtilizationOfUnsecuredLines
x <= 3.0m	0.025000	0.100000
3.0m < x <= 19.2m	0.012000	0.099000
19.2m < x <= 43.6m	0.014000	0.103000
43.6m < x <= 83.8m	0.019000	0.103000
83.8m < x <= 155.4m	0.022000	0.100000
155.4m < x <= 273.6m	0.031000	0.100000
273.6m < x <= 447.4m	0.052000	0.099000
447.4m < x <= 700.7m	0.090000	0.098000
700.7m < x	0.199000	0.198000

Grouping modalities   : 100%|██████████| 162/162 [00:00<00:00, 6461.12it/s]
Computing associations: 100%|██████████| 162/162 [00:00<00:00, 3819.84it/s]
Testing robustness    :   0%|          | 0/162 [00:00<?, ?it/s]

 - [AutoCarver] Carved feature distribution

X distribution

	target_rate	frequency
x <= 3.0m	0.020000	0.500000
155.4m < x <= 273.6m	0.037000	0.100000
273.6m < x <= 447.4m	0.053000	0.100000
447.4m < x <= 700.7m	0.088000	0.100000
700.7m < x	0.199000	0.200000

X_dev distribution

	target_rate	frequency
x <= 3.0m	0.018000	0.504000
155.4m < x <= 273.6m	0.031000	0.100000
273.6m < x <= 447.4m	0.052000	0.099000
447.4m < x <= 700.7m	0.090000	0.098000
700.7m < x	0.199000	0.198000

------


------
[AutoCarver] Fit NumberOfTimes90DaysLate (5/10)
---

 - [AutoCarver] Raw feature distribution

X distribution

	target_rate	frequency
NumberOfTimes90DaysLate
x <= nan	0.067000	1.000000

X_dev distribution

	target_rate	frequency
NumberOfTimes90DaysLate
x <= nan	0.066000	1.000000

 - [AutoCarver] No robust combination for feature 'NumberOfTimes90DaysLate' could be found. It will be ignored. You might have to increase the size of your test sample (test sample not representative of test sample for this feature) or you should consider dropping this features.
------


------
[AutoCarver] Fit NumberOfTime60-89DaysPastDueNotWorse (6/10)
---

 - [AutoCarver] Raw feature distribution

X distribution

	target_rate	frequency
NumberOfTime60-89DaysPastDueNotWorse
x <= nan	0.067000	1.000000

X_dev distribution

	target_rate	frequency
NumberOfTime60-89DaysPastDueNotWorse
x <= nan	0.066000	1.000000

 - [AutoCarver] No robust combination for feature 'NumberOfTime60-89DaysPastDueNotWorse' could be found. It will be ignored. You might have to increase the size of your test sample (test sample not representative of test sample for this feature) or you should consider dropping this features.
------


------
[AutoCarver] Fit NumberOfOpenCreditLinesAndLoans (7/10)
---

 - [AutoCarver] Raw feature distribution

X distribution

	target_rate	frequency
NumberOfOpenCreditLinesAndLoans
x <= 3.0	0.107000	0.147000
3.0 < x <= 5.0	0.064000	0.163000
5.0 < x <= 8.0	0.053000	0.262000
8.0 < x <= 10.0	0.061000	0.141000
10.0 < x <= 12.0	0.061000	0.102000
12.0 < x	0.067000	0.185000

X_dev distribution

	target_rate	frequency
NumberOfOpenCreditLinesAndLoans
x <= 3.0	0.107000	0.147000
3.0 < x <= 5.0	0.063000	0.164000
5.0 < x <= 8.0	0.053000	0.265000
8.0 < x <= 10.0	0.055000	0.138000
10.0 < x <= 12.0	0.057000	0.102000
12.0 < x	0.066000	0.183000

Grouping modalities   : 100%|██████████| 30/30 [00:00<00:00, 3600.36it/s]
Computing associations: 100%|██████████| 30/30 [00:00<00:00, 3562.24it/s]
Testing robustness    :   0%|          | 0/30 [00:00<?, ?it/s]

 - [AutoCarver] Carved feature distribution

X distribution

	target_rate	frequency
x <= 3.0	0.107000	0.147000
3.0 < x <= 5.0	0.064000	0.163000
5.0 < x <= 8.0	0.053000	0.262000
8.0 < x <= 10.0	0.061000	0.243000
12.0 < x	0.067000	0.185000

X_dev distribution

	target_rate	frequency
x <= 3.0	0.107000	0.147000
3.0 < x <= 5.0	0.063000	0.164000
5.0 < x <= 8.0	0.053000	0.265000
8.0 < x <= 10.0	0.056000	0.240000
12.0 < x	0.066000	0.183000

------


------
[AutoCarver] Fit NumberRealEstateLoansOrLines (8/10)
---

 - [AutoCarver] Raw feature distribution

X distribution

	target_rate	frequency
NumberRealEstateLoansOrLines
x <= 0.0	0.083000	0.376000
0.0 < x <= 1.0	0.053000	0.348000
1.0 < x	0.065000	0.277000

X_dev distribution

	target_rate	frequency
NumberRealEstateLoansOrLines
x <= 0.0	0.083000	0.373000
0.0 < x <= 1.0	0.052000	0.352000
1.0 < x	0.059000	0.276000

Grouping modalities   : 100%|██████████| 3/3 [00:00<00:00, 1506.03it/s]
Computing associations: 100%|██████████| 3/3 [00:00<00:00, 3008.83it/s]
Testing robustness    :   0%|          | 0/3 [00:00<?, ?it/s]

 - [AutoCarver] Carved feature distribution

X distribution

	target_rate	frequency
x <= 0.0	0.083000	0.376000
0.0 < x <= 1.0	0.053000	0.348000
1.0 < x	0.065000	0.277000

X_dev distribution

	target_rate	frequency
x <= 0.0	0.083000	0.373000
0.0 < x <= 1.0	0.052000	0.352000
1.0 < x	0.059000	0.276000

------


------
[AutoCarver] Fit age (9/10)
---

 - [AutoCarver] Raw feature distribution

X distribution

	target_rate	frequency
age
x <= 33.0	0.115000	0.114000
33.0 < x <= 39.0	0.097000	0.100000
39.0 < x <= 48.0	0.085000	0.203000
48.0 < x <= 56.0	0.071000	0.193000
56.0 < x <= 61.0	0.051000	0.112000
61.0 < x	0.028000	0.277000

X_dev distribution

	target_rate	frequency
age
x <= 33.0	0.110000	0.114000
33.0 < x <= 39.0	0.094000	0.098000
39.0 < x <= 48.0	0.082000	0.204000
48.0 < x <= 56.0	0.072000	0.194000
56.0 < x <= 61.0	0.048000	0.113000
61.0 < x	0.029000	0.277000

Grouping modalities   : 100%|██████████| 30/30 [00:00<00:00, 3313.21it/s]
Computing associations: 100%|██████████| 30/30 [00:00<00:00, 2360.64it/s]
Testing robustness    :   0%|          | 0/30 [00:00<?, ?it/s]

 - [AutoCarver] Carved feature distribution

X distribution

	target_rate	frequency
x <= 33.0	0.115000	0.114000
33.0 < x <= 39.0	0.089000	0.304000
48.0 < x <= 56.0	0.071000	0.193000
56.0 < x <= 61.0	0.051000	0.112000
61.0 < x	0.028000	0.277000

X_dev distribution

	target_rate	frequency
x <= 33.0	0.110000	0.114000
33.0 < x <= 39.0	0.086000	0.302000
48.0 < x <= 56.0	0.072000	0.194000
56.0 < x <= 61.0	0.048000	0.113000
61.0 < x	0.029000	0.277000

------


------
[AutoCarver] Fit MonthlyIncome (10/10)
---

 - [AutoCarver] Raw feature distribution

X distribution

	target_rate	frequency
MonthlyIncome
x <= 2.3K	0.085000	0.101000
2.3K < x <= 3.4K	0.097000	0.101000
3.4K < x <= 5.4K	0.080000	0.201000
5.4K < x <= 8.2K	0.061000	0.199000
8.2K < x <= 10.7K	0.051000	0.100000
10.7K < x	0.044000	0.100000
__NAN__	0.056000	0.197000

X_dev distribution

	target_rate	frequency
MonthlyIncome
x <= 2.3K	0.089000	0.101000
2.3K < x <= 3.4K	0.098000	0.101000
3.4K < x <= 5.4K	0.075000	0.199000
5.4K < x <= 8.2K	0.060000	0.199000
8.2K < x <= 10.7K	0.042000	0.098000
10.7K < x	0.047000	0.102000
__NAN__	0.056000	0.200000

Grouping modalities   : 100%|██████████| 30/30 [00:00<00:00, 3022.63it/s]
Computing associations: 100%|██████████| 30/30 [00:00<00:00, 2698.00it/s]
Testing robustness    :   0%|          | 0/30 [00:00<?, ?it/s]

 - [AutoCarver] Carved feature distribution

X distribution

	target_rate	frequency
x <= 2.3K	0.085000	0.101000
2.3K < x <= 3.4K	0.097000	0.101000
3.4K < x <= 5.4K	0.080000	0.201000
5.4K < x <= 8.2K	0.061000	0.199000
8.2K < x <= 10.7K	0.047000	0.201000
__NAN__	0.056000	0.197000

X_dev distribution

	target_rate	frequency
x <= 2.3K	0.089000	0.101000
2.3K < x <= 3.4K	0.098000	0.101000
3.4K < x <= 5.4K	0.075000	0.199000
5.4K < x <= 8.2K	0.060000	0.199000
8.2K < x <= 10.7K	0.044000	0.200000
__NAN__	0.056000	0.200000

------

 - [BaseDiscretizer] Transform Quantitative ['age', 'NumberOfDependents', 'DebtRatio', 'RevolvingUtilizationOfUnsecuredLines', 'NumberOfOpenCreditLinesAndLoans', 'NumberRealEstateLoansOrLines', 'NumberOfTime30-59DaysPastDueNotWorse', 'MonthlyIncome']

Inspecting Discretization¶

In [7]:

x_discretized[quantitative_features].head()

Out[7]:

	RevolvingUtilizationOfUnsecuredLines	age	NumberOfTime30-59DaysPastDueNotWorse	DebtRatio	MonthlyIncome	NumberOfOpenCreditLinesAndLoans	NumberOfTimes90DaysLate	NumberRealEstateLoansOrLines	NumberOfTime60-89DaysPastDueNotWorse	NumberOfDependents
87936	0.0	4	0	4.0	NaN	2	0	1	0	0.0
3893	3.0	0	0	2.0	2.0	1	0	0	0	0.0
41405	4.0	1	0	4.0	NaN	2	0	1	0	0.0
91125	2.0	2	0	2.0	3.0	3	0	0	0	2.0
67373	3.0	2	3	4.0	NaN	2	2	1	1	NaN

In [7]:

auto_carver.summary()

Out[7]:

		label	content
feature	dtype
DebtRatio	float	0	[x <= 287.6m]
	float	1	[287.6m < x <= 467.4m]
	float	2	[467.4m < x <= 648.0m]
	float	3	[648.0m < x <= 3.8]
	float	4	[3.8 < x]
MonthlyIncome	float	0	[x <= 2.3K]
	float	1	[2.3K < x <= 3.4K]
	float	2	[3.4K < x <= 5.4K]
	float	3	[5.4K < x <= 8.2K]
	float	4	[8.2K < x]
	float	5	[__NAN__]
NumberOfDependents	float	0	[x <= 0.0]
	float	1	[0.0 < x <= 1.0]
	float	2	[1.0 < x]
	float	3	[__NAN__]
NumberOfOpenCreditLinesAndLoans	float	0	[x <= 3.0]
	float	1	[3.0 < x <= 5.0]
	float	2	[5.0 < x <= 8.0]
	float	3	[8.0 < x <= 12.0]
	float	4	[12.0 < x]
NumberOfTime30-59DaysPastDueNotWorse	float	0	[x <= 0.0]
	float	1	[0.0 < x]
NumberRealEstateLoansOrLines	float	0	[x <= 0.0]
	float	1	[0.0 < x <= 1.0]
	float	2	[1.0 < x]
RevolvingUtilizationOfUnsecuredLines	float	0	[x <= 155.4m]
	float	1	[155.4m < x <= 273.6m]
	float	2	[273.6m < x <= 447.4m]
	float	3	[447.4m < x <= 700.7m]
	float	4	[700.7m < x]
age	float	0	[x <= 33.0]
	float	1	[33.0 < x <= 48.0]
	float	2	[48.0 < x <= 56.0]
	float	3	[56.0 < x <= 61.0]
	float	4	[61.0 < x]

Saving for later uses¶

In [9]:

import json

# storing as json file
with open('my_carver.json', 'w') as my_carver_json:
    json.dump(auto_carver.to_json(), my_carver_json)

Feature Selection¶

Setting up measures and filters¶

In [10]:

from AutoCarver.feature_selection import FeatureSelector

n_best = 10  # number of features to select

feature_selector = FeatureSelector(
    quantitative_features=quantitative_features, 
    n_best=n_best,
    pretty_print=True,
)
best_features = feature_selector.select(x_discretized, x_discretized[target])

------
[FeatureSelector] Selecting from Features: ['age', 'NumberOfDependents', 'DebtRatio', 'RevolvingUtilizationOfUnsecuredLines', 'NumberOfTimes90DaysLate', 'NumberOfTime60-89DaysPastDueNotWorse', 'NumberOfOpenCreditLinesAndLoans', 'NumberRealEstateLoansOrLines', 'NumberOfTime30-59DaysPastDueNotWorse', 'MonthlyIncome']
---

 - Association between X and y

	dtype	pct_nan	pct_mode	mode	kruskal_measure
NumberOfTimes90DaysLate	int64	0.000000	0.943960	0	11854.250713
NumberOfTime60-89DaysPastDueNotWorse	int64	0.000000	0.949254	0	7636.377475
RevolvingUtilizationOfUnsecuredLines	float64	0.000000	0.500000	0.000000	6158.299923
NumberOfTime30-59DaysPastDueNotWorse	int64	0.000000	0.839035	0	6076.412797
age	int64	0.000000	0.303592	1	1370.032527
MonthlyIncome	float64	0.197383	0.200945	2.000000	337.382149
NumberOfDependents	float64	0.026129	0.579741	0.000000	175.384642
NumberOfOpenCreditLinesAndLoans	int64	0.000000	0.261602	2	121.575254
NumberRealEstateLoansOrLines	int64	0.000000	0.375512	0	120.839911
DebtRatio	float64	0.000000	0.400000	0.000000	47.123007

 - Association between X and y, filtered for inter-feature assocation

	dtype	pct_nan	pct_mode	mode	kruskal_measure
NumberOfTimes90DaysLate	int64	0.000000	0.943960	0	11854.250713
NumberOfTime60-89DaysPastDueNotWorse	int64	0.000000	0.949254	0	7636.377475
RevolvingUtilizationOfUnsecuredLines	float64	0.000000	0.500000	0.000000	6158.299923
NumberOfTime30-59DaysPastDueNotWorse	int64	0.000000	0.839035	0	6076.412797
age	int64	0.000000	0.303592	1	1370.032527
MonthlyIncome	float64	0.197383	0.200945	2.000000	337.382149
NumberOfDependents	float64	0.026129	0.579741	0.000000	175.384642
NumberOfOpenCreditLinesAndLoans	int64	0.000000	0.261602	2	121.575254
NumberRealEstateLoansOrLines	int64	0.000000	0.375512	0	120.839911
DebtRatio	float64	0.000000	0.400000	0.000000	47.123007

------

Out[10]:

['NumberOfTimes90DaysLate',
 'NumberOfTime60-89DaysPastDueNotWorse',
 'RevolvingUtilizationOfUnsecuredLines',
 'NumberOfTime30-59DaysPastDueNotWorse',
 'age',
 'MonthlyIncome',
 'NumberOfDependents',
 'NumberOfOpenCreditLinesAndLoans',
 'NumberRealEstateLoansOrLines',
 'DebtRatio']

Enjoy modeling!¶