CostSensitiveLearning

This repository contains the code for the papers:

1. "Predict-then-optimize or predict-and-optimize? An empirical evaluation of cost-sensitive learning strategies" Information Sciences (2022).
2. "A new perspective on classification: Optimally allocating limited resources to uncertain tasks" Decision Support Systems (2023).

For (1), an experiment is conducted by running the experiments/overview.py file where settings, data set, methodologies, and evaluators can be chosen. For (2), experiments can similarly be replicated using experiments/overview_ranking.py.

Data

Due to size limitations, not all data is provided in this repository. Data sets can be found online at the following links:

• Kaggle Credit Card Fraud
• Kaggle IEEE Fraud Detection
• UCI KDD98 Direct Mailing
• UCI Bank Marketing
• Kaggle Telco Customer Churn
• TV Subscription Churn
• Kaggle Give Me Some Credit
• UCI Default of Credit Card Clients
• VUB Credit Scoring

Acknowledgments

The code for cslogit and csboost are Python versions of the original cslogit by Sebastiaan Höppner et al.. Thank you to Lennert Van der Schraelen for sharing his Python adaptation!

Reference

If you use this software, please cite it as follows:

@article{vanderschueren2022predict,
  title={Predict-then-optimize or predict-and-optimize? An empirical evaluation of cost-sensitive learning strategies},
  author={Vanderschueren, Toon and Verdonck, Tim and Baesens, Bart and Verbeke, Wouter},
  journal={Information Sciences},
  volume={594},
  pages={400--415},
  year={2022},
  publisher={Elsevier}
}

@article{vanderschueren2023new,
  title={A new perspective on classification: Optimally allocating limited resources to uncertain tasks},
  author={Vanderschueren, Toon and Baesens, Bart and Verdonck, Tim and Verbeke, Wouter},
  journal={Decision Support Systems},
  pages={114151},
  year={2023},
  publisher={Elsevier}
}

Contact the corresponding author at toon.vanderschueren@gmail.com.

All content in this repository is licensed under the MIT license.