load_churn_tv_subscriptions

empulse.datasets.load_churn_tv_subscriptions(*, backend)

Load the TV Subscription Churn dataset (binary classification).

The goal is to predict whether a customer will churn or not. The target variable is whether the customer churned, ‘yes’ = 1 and ‘no’ = 0.

This dataset is from a TV cable provider containing all 9410 customers active during the first semester of 2014. Feature names are anonymized to protect the privacy of the customers.

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

Classes	2
Churners	455
Non-churners	8955
Samples	9410
Features	45

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 the precomputed per-customer costs: tp_cost, fp_cost, tn_cost, fn_cost.

Notes

The per-instance costs stored in the dataset were derived from the churn retention cost model of Bahnsen et al. (2015). The individual cost drivers (CLV, incentive cost, contact cost, acceptance probability) are not stored; only the final computed values are available.

References

[1]

A. Correa Bahnsen, D.Aouada, B, Ottersten, “A novel cost-sensitive framework for customer churn predictive modeling”, Decision Analytics, 2:5, 2015.

Examples

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

dataset = load_churn_tv_subscriptions(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)