load_vub_credit_scoring#

empulse.datasets.load_vub_credit_scoring(*, backend)[source]#

Load the VUB Credit Scoring dataset (binary classification).

The goal is to predict whether a borrower will experience 45+ days of payment delay (default). The target variable is whether the borrower defaulted, 1 = default, 0 = no default.

This dataset is from a Romanian non-banking financial institution (NBFI) provided by the VUB Data Analytics Laboratory (Petrides et al., 2020).

For additional information about the dataset, consult the User Guide.

Classes

2

Defaulters

3206

Non-defaulters

15711

Samples

18917

Features

16
Parameters:
backendmodule

Dataframe library to use for data and target. Pass the library module directly, e.g. backend=polars or backend=pandas.

Returns:
datasetDataset

instance_costs contains:

  • 'cl': shifted loan amount per borrower.

  • 'fp_cost': precomputed FP cost per borrower following Bahnsen et al. (2014).

Notes

Cost matrix (Bahnsen et al. 2014, Petrides et al. 2020, Vanderschueren et al. 2022):

Actual positive \(y_i = 1\)

Actual negative \(y_i = 0\)

Predicted positive \(\hat{y}_i = 1\)

tp_cost \(= 0\)

fp_cost (precomputed per borrower)

Predicted negative \(\hat{y}_i = 0\)

fn_cost \(= Cl_i \cdot L_{gd}\)

tn_cost \(= 0\)

The cost matrix uses symbolic parameters with the following defaults:

  • loss_given_default (\(L_{gd}\)) = 0.75

The false positive cost is precomputed with an annual interest rate of 4.79%, an annual cost of funds of 2.94% and a 24-month term.

The published data is anonymised: every monetary column, including the loan amount, is standardised to zero mean and unit variance. The original costs of Petrides et al. (2020) therefore cannot be recovered. Following Vanderschueren et al. (2022), the loan amount is shifted to be strictly positive (\(Cl_i = Loan\_amount_i - \min_j Loan\_amount_j + 10^{-9}\)) and used as the credit line of the Bahnsen et al. (2014) cost matrix. The costs are therefore in arbitrary units: only their relative size is meaningful.

References

[1]

Petrides, G., Moldovan, D., Coenen, L., Guns, T., & Verbeke, W. (2020). Cost-sensitive learning for profit-driven credit scoring. Journal of the Operational Research Society, 1–13.

[2]

Vanderschueren, T., Verdonck, T., Baesens, B., & Verbeke, W. (2022). Predict-then-optimize or predict-and-optimize? An empirical evaluation of cost-sensitive learning strategies. Information Sciences, 594, 400–415.

[3]

Bahnsen, A. C., Aouada, D., & Ottersten, B. (2014). Example-dependent cost-sensitive logistic regression for credit scoring. In 2014 13th International Conference on Machine Learning and Applications (pp. 263–269).

Examples

import numpy as np
import pandas as pd
from empulse.datasets import load_vub_credit_scoring
from empulse.metrics import Metric, Cost

dataset = load_vub_credit_scoring(backend=pd)

# replace with your own model's predicted probabilities
y_score = np.random.default_rng(0).uniform(size=len(dataset.target))

metric = Metric(dataset.cost_matrix, Cost())
score = metric(dataset.target, y_score, **dataset.instance_costs)